feat: add OAuth token authentication support (#236)

* feat: add OAuth token authentication support

Add claude_code_oauth_token as an alternative authentication method to anthropic_api_key.
This provides more flexibility for users who prefer OAuth authentication.

- Add claude_code_oauth_token input to action.yml
- Pass OAuth token through to claude-code-base-action
- Update README with OAuth token documentation and examples
- Update security best practices to cover both authentication methods
- Add OAuth example to examples/claude.yml

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* docs: add OAuth token generation instructions for Pro/Max users

Update README to mention that Pro and Max users can generate OAuth tokens
by running `claude setup-token` locally. This provides clearer guidance
for users who want to use OAuth authentication instead of API keys.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* docs: update CI capabilities documentation

- Move GitHub Actions access from limitations to capabilities in README
- Update FAQ to explain how to enable CI/CD access with actions:read permission
- Clarify that Claude can access workflow results on PRs where it's tagged

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

---------

Co-authored-by: Claude <noreply@anthropic.com>

.github/workflows/issue-triage.yml

@@ -32,7 +32,7 @@ jobs:
                   "--rm",
                   "-e",
                   "GITHUB_PERSONAL_ACCESS_TOKEN",
-                  "ghcr.io/github/github-mcp-server:sha-6d69797"
+                  "ghcr.io/github/github-mcp-server:sha-721fd3e"
                 ],
                 "env": {
                   "GITHUB_PERSONAL_ACCESS_TOKEN": "${{ secrets.GITHUB_TOKEN }}"

.prettierignore

@@ -0,0 +1,2 @@
+# Test fixtures should not be formatted to preserve exact output matching
+test/fixtures/

CONTRIBUTING.md

@@ -50,20 +50,6 @@ Thank you for your interest in contributing to Claude Code Action! This document
    bun test
    ```
 
-2. **Integration Tests** (using GitHub Actions locally):
-
-   ```bash
-   ./test-local.sh
-   ```
-
-   This script:
-
-   - Installs `act` if not present (requires Homebrew on macOS)
-   - Runs the GitHub Action workflow locally using Docker
-   - Requires your `ANTHROPIC_API_KEY` to be set
-
-   On Apple Silicon Macs, the script automatically adds the `--container-architecture linux/amd64` flag to avoid compatibility issues.
-
 ## Pull Request Process
 
 1. Create a new branch from `main`:
@@ -103,13 +89,7 @@ Thank you for your interest in contributing to Claude Code Action! This document
 
 When modifying the action:
 
-1. Test locally with the test script:
+1. Test in a real GitHub Actions workflow by:
-
-   ```bash
-   ./test-local.sh
-   ```
-
-2. Test in a real GitHub Actions workflow by:
    - Creating a test repository
    - Using your branch as the action source:
      ```yaml

FAQ.md

@@ -12,6 +12,10 @@ The `github-actions` user cannot trigger subsequent GitHub Actions workflows. Th
 
 Only users with **write permissions** to the repository can trigger Claude. This is a security feature to prevent unauthorized use. Make sure the user commenting has at least write access to the repository.
 
+### Why can't I assign @claude to an issue on my repository?
+
+If you're in a public repository, you should be able to assign to Claude without issue. If it's a private organization repository, you can only assign to users in your own organization, which Claude isn't. In this case, you'll need to make a custom user in that case.
+
 ### Why am I getting OIDC authentication errors?
 
 If you're using the default GitHub App authentication, you must add the `id-token: write` permission to your workflow:
@@ -47,14 +51,29 @@ allowed_tools: "Bash(git rebase:*)" # Use with caution
 
 Claude doesn't create PRs by default. Instead, it pushes commits to a branch and provides a link to a pre-filled PR submission page. This approach ensures your repository's branch protection rules are still adhered to and gives you final control over PR creation.
 
-### Why can't Claude run my tests or see CI results?
+### Can Claude see my GitHub Actions CI results?
 
-Claude cannot access GitHub Actions logs, test results, or other CI/CD outputs by default. It only has access to the repository files. If you need Claude to see test results, you can either:
+Yes! Claude can access GitHub Actions workflow runs, job logs, and test results on the PR where it's tagged. To enable this:
 
-1. Instruct Claude to run tests before making commits
+1. Add `actions: read` permission to your workflow:
-2. Copy and paste CI results into a comment for Claude to analyze
 
-This limitation exists for security reasons but may be reconsidered in the future based on user feedback.
+   ```yaml
+   permissions:
+     contents: write
+     pull-requests: write
+     issues: write
+     actions: read
+   ```
+
+2. Configure the action with additional permissions:
+   ```yaml
+   - uses: anthropics/claude-code-action@beta
+     with:
+       additional_permissions: |
+         actions: read
+   ```
+
+Claude will then be able to analyze CI failures and help debug workflow issues. For running tests locally before commits, you can still instruct Claude to do so in your request.
 
 ### Why does Claude only update one comment instead of creating new ones?
 

README.md

@@ -30,7 +30,9 @@ This command will guide you through setting up the GitHub app and required secre
 **Requirements**: You must be a repository admin to complete these steps.
 
 1. Install the Claude GitHub app to your repository: https://github.com/apps/claude
-2. Add `ANTHROPIC_API_KEY` to your repository secrets ([Learn how to use secrets in GitHub Actions](https://docs.github.com/en/actions/security-for-github-actions/security-guides/using-secrets-in-github-actions))
+2. Add authentication to your repository secrets ([Learn how to use secrets in GitHub Actions](https://docs.github.com/en/actions/security-for-github-actions/security-guides/using-secrets-in-github-actions)):
+   - Either `ANTHROPIC_API_KEY` for API key authentication
+   - Or `CLAUDE_CODE_OAUTH_TOKEN` for OAuth token authentication (Pro and Max users can generate this by running `claude setup-token` locally)
 3. Copy the workflow file from [`examples/claude.yml`](./examples/claude.yml) into your repository's `.github/workflows/`
 
 ## 📚 FAQ
@@ -49,7 +51,7 @@ on:
   pull_request_review_comment:
     types: [created]
   issues:
-    types: [opened, assigned]
+    types: [opened, assigned, labeled]
   pull_request_review:
     types: [submitted]
 
@@ -60,11 +62,15 @@ jobs:
       - uses: anthropics/claude-code-action@beta
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          # Or use OAuth token instead:
+          # claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
           github_token: ${{ secrets.GITHUB_TOKEN }}
           # Optional: add custom trigger phrase (default: @claude)
           # trigger_phrase: "/claude"
           # Optional: add assignee trigger for issues
           # assignee_trigger: "claude"
+          # Optional: add label trigger for issues
+          # label_trigger: "claude"
           # Optional: add custom environment variables (YAML format)
           # claude_env: |
           #   NODE_ENV: test
@@ -72,18 +78,25 @@ jobs:
           #   API_URL: https://api.example.com
           # Optional: limit the number of conversation turns
           # max_turns: "5"
+          # Optional: grant additional permissions (requires corresponding GitHub token permissions)
+          # additional_permissions: |
+          #   actions: read
 ```
 
 ## Inputs
 
 | Input                     | Description                                                                                                          | Required | Default   |
-| --------------------- | -------------------------------------------------------------------------------------------------------------------- | -------- | --------- |
+| ------------------------- | -------------------------------------------------------------------------------------------------------------------- | -------- | --------- |
 | `anthropic_api_key`       | Anthropic API key (required for direct API, not needed for Bedrock/Vertex)                                           | No\*     | -         |
+| `claude_code_oauth_token` | Claude Code OAuth token (alternative to anthropic_api_key)                                                           | No\*     | -         |
 | `direct_prompt`           | Direct prompt for Claude to execute automatically without needing a trigger (for automated workflows)                | No       | -         |
+| `base_branch`             | The base branch to use for creating new branches (e.g., 'main', 'develop')                                           | No       | -         |
 | `max_turns`               | Maximum number of conversation turns Claude can take (limits back-and-forth exchanges)                               | No       | -         |
 | `timeout_minutes`         | Timeout in minutes for execution                                                                                     | No       | `30`      |
+| `use_sticky_comment`      | Use just one comment to deliver PR comments (only applies for pull_request event workflows)                          | No       | `false`   |
 | `github_token`            | GitHub token for Claude to operate with. **Only include this if you're connecting a custom GitHub app of your own!** | No       | -         |
 | `model`                   | Model to use (provider-specific format required for Bedrock/Vertex)                                                  | No       | -         |
+| `fallback_model`          | Enable automatic fallback to specified model when primary model is unavailable                                       | No       | -         |
 | `anthropic_model`         | **DEPRECATED**: Use `model` instead. Kept for backward compatibility.                                                | No       | -         |
 | `use_bedrock`             | Use Amazon Bedrock with OIDC authentication instead of direct Anthropic API                                          | No       | `false`   |
 | `use_vertex`              | Use Google Vertex AI with OIDC authentication instead of direct Anthropic API                                        | No       | `false`   |
@@ -92,8 +105,11 @@ jobs:
 | `custom_instructions`     | Additional custom instructions to include in the prompt for Claude                                                   | No       | ""        |
 | `mcp_config`              | Additional MCP configuration (JSON string) that merges with the built-in GitHub MCP servers                          | No       | ""        |
 | `assignee_trigger`        | The assignee username that triggers the action (e.g. @claude). Only used for issue assignment                        | No       | -         |
+| `label_trigger`           | The label name that triggers the action when applied to an issue (e.g. "claude")                                     | No       | -         |
 | `trigger_phrase`          | The trigger phrase to look for in comments, issue/PR bodies, and issue titles                                        | No       | `@claude` |
+| `branch_prefix`           | The prefix to use for Claude branches (defaults to 'claude/', use 'claude-' for dash format)                         | No       | `claude/` |
 | `claude_env`              | Custom environment variables to pass to Claude Code execution (YAML format)                                          | No       | ""        |
+| `additional_permissions`  | Additional permissions to enable. Currently supports 'actions: read' for viewing workflow results                    | No       | ""        |
 
 \*Required when using direct Anthropic API (default and when not using Bedrock or Vertex)
 
@@ -319,6 +335,7 @@ This action is built on top of [`anthropics/claude-code-base-action`](https://gi
   - When triggered on an **issue**: Always creates a new branch for the work
   - When triggered on an **open PR**: Always pushes directly to the existing PR branch
   - When triggered on a **closed PR**: Creates a new branch since the original is no longer active
+- **View GitHub Actions Results**: Can access workflow runs, job logs, and test results on the PR where it's tagged when `actions: read` permission is configured (see [Additional Permissions for CI/CD Integration](#additional-permissions-for-cicd-integration))
 
 ### What Claude Cannot Do
 
@@ -327,11 +344,79 @@ This action is built on top of [`anthropics/claude-code-base-action`](https://gi
 - **Post Multiple Comments**: Claude only acts by updating its initial comment
 - **Execute Commands Outside Its Context**: Claude only has access to the repository and PR/issue context it's triggered in
 - **Run Arbitrary Bash Commands**: By default, Claude cannot execute Bash commands unless explicitly allowed using the `allowed_tools` configuration
-- **View CI/CD Results**: Cannot access CI systems, test results, or build logs unless an additional tool or MCP server is configured
 - **Perform Branch Operations**: Cannot merge branches, rebase, or perform other git operations beyond pushing commits
 
 ## Advanced Configuration
 
+### Additional Permissions for CI/CD Integration
+
+The `additional_permissions` input allows Claude to access GitHub Actions workflow information when you grant the necessary permissions. This is particularly useful for analyzing CI/CD failures and debugging workflow issues.
+
+#### Enabling GitHub Actions Access
+
+To allow Claude to view workflow run results, job logs, and CI status:
+
+1. **Grant the necessary permission to your GitHub token**:
+
+   - When using the default `GITHUB_TOKEN`, add the `actions: read` permission to your workflow:
+
+   ```yaml
+   permissions:
+     contents: write
+     pull-requests: write
+     issues: write
+     actions: read # Add this line
+   ```
+
+2. **Configure the action with additional permissions**:
+
+   ```yaml
+   - uses: anthropics/claude-code-action@beta
+     with:
+       anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+       additional_permissions: |
+         actions: read
+       # ... other inputs
+   ```
+
+3. **Claude will automatically get access to CI/CD tools**:
+   When you enable `actions: read`, Claude can use the following MCP tools:
+   - `mcp__github_ci__get_ci_status` - View workflow run statuses
+   - `mcp__github_ci__get_workflow_run_details` - Get detailed workflow information
+   - `mcp__github_ci__download_job_log` - Download and analyze job logs
+
+#### Example: Debugging Failed CI Runs
+
+```yaml
+name: Claude CI Helper
+on:
+  issue_comment:
+    types: [created]
+
+permissions:
+  contents: write
+  pull-requests: write
+  issues: write
+  actions: read # Required for CI access
+
+jobs:
+  claude-ci-helper:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: anthropics/claude-code-action@beta
+        with:
+          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          additional_permissions: |
+            actions: read
+          # Now Claude can respond to "@claude why did the CI fail?"
+```
+
+**Important Notes**:
+
+- The GitHub token must have the `actions: read` permission in your workflow
+- If the permission is missing, Claude will warn you and suggest adding it
+- Currently, only `actions: read` is supported, but the format allows for future extensions
+
 ### Custom Environment Variables
 
 You can pass custom environment variables to Claude Code execution using the `claude_env` input. This is useful for CI/test setups that require specific environment variables:
@@ -524,18 +609,21 @@ The [Claude Code GitHub app](https://github.com/apps/claude) requires these perm
 
 All commits made by Claude through this action are automatically signed with commit signatures. This ensures the authenticity and integrity of commits, providing a verifiable trail of changes made by the action.
 
-### ⚠️ ANTHROPIC_API_KEY Protection
+### ⚠️ Authentication Protection
 
-**CRITICAL: Never hardcode your Anthropic API key in workflow files!**
+**CRITICAL: Never hardcode your Anthropic API key or OAuth token in workflow files!**
 
-Your ANTHROPIC_API_KEY must always be stored in GitHub secrets to prevent unauthorized access:
+Your authentication credentials must always be stored in GitHub secrets to prevent unauthorized access:
 
 ```yaml
 # CORRECT ✅
 anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+# OR
+claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
 
 # NEVER DO THIS ❌
 anthropic_api_key: "sk-ant-api03-..." # Exposed and vulnerable!
+claude_code_oauth_token: "oauth_token_..." # Exposed and vulnerable!
 ```
 
 ### Setting Up GitHub Secrets
@@ -543,17 +631,18 @@ anthropic_api_key: "sk-ant-api03-..." # Exposed and vulnerable!
 1. Go to your repository's Settings
 2. Click on "Secrets and variables" → "Actions"
 3. Click "New repository secret"
-4. Name: `ANTHROPIC_API_KEY`
+4. For authentication, choose one:
-5. Value: Your Anthropic API key (starting with `sk-ant-`)
+   - API Key: Name: `ANTHROPIC_API_KEY`, Value: Your Anthropic API key (starting with `sk-ant-`)
-6. Click "Add secret"
+   - OAuth Token: Name: `CLAUDE_CODE_OAUTH_TOKEN`, Value: Your Claude Code OAuth token (Pro and Max users can generate this by running `claude setup-token` locally)
+5. Click "Add secret"
 
-### Best Practices for ANTHROPIC_API_KEY
+### Best Practices for Authentication
 
-1. ✅ Always use `${{ secrets.ANTHROPIC_API_KEY }}` in workflows
+1. ✅ Always use `${{ secrets.ANTHROPIC_API_KEY }}` or `${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}` in workflows
-2. ✅ Never commit API keys to version control
+2. ✅ Never commit API keys or tokens to version control
-3. ✅ Regularly rotate your API keys
+3. ✅ Regularly rotate your API keys and tokens
 4. ✅ Use environment secrets for organization-wide access
-5. ❌ Never share API keys in pull requests or issues
+5. ❌ Never share API keys or tokens in pull requests or issues
 6. ❌ Avoid logging workflow variables that might contain keys
 
 ## Security Best Practices

ROADMAP.md

@@ -4,7 +4,7 @@ Thank you for trying out the beta of our GitHub Action! This document outlines o
 
 ## Path to 1.0
 
-- **Ability to see GitHub Action CI results** - This will enable Claude to look at CI failures and make updates to PRs to fix test failures, lint errors, and the like.
+- ~**Ability to see GitHub Action CI results** - This will enable Claude to look at CI failures and make updates to PRs to fix test failures, lint errors, and the like.~
 - **Cross-repo support** - Enable Claude to work across multiple repositories in a single session
 - **Ability to modify workflow files** - Let Claude update GitHub Actions workflows and other CI configuration files
 - **Support for workflow_dispatch and repository_dispatch events** - Dispatch Claude on events triggered via API from other workflows or from other services

action.yml

@@ -12,9 +12,17 @@ inputs:
   assignee_trigger:
     description: "The assignee username that triggers the action (e.g. @claude)"
     required: false
+  label_trigger:
+    description: "The label that triggers the action (e.g. claude)"
+    required: false
+    default: "claude"
   base_branch:
     description: "The branch to use as the base/source when creating new branches (defaults to repository default branch)"
     required: false
+  branch_prefix:
+    description: "The prefix to use for Claude branches (defaults to 'claude/', use 'claude-' for dash format)"
+    required: false
+    default: "claude/"
 
   # Claude Code configuration
   model:
@@ -23,6 +31,9 @@ inputs:
   anthropic_model:
     description: "DEPRECATED: Use 'model' instead. Model to use (provider-specific format required for Bedrock/Vertex)"
     required: false
+  fallback_model:
+    description: "Enable automatic fallback to specified model when primary model is unavailable"
+    required: false
   allowed_tools:
     description: "Additional tools for Claude to use (the base GitHub tools will always be included)"
     required: false
@@ -41,6 +52,10 @@ inputs:
     default: ""
   mcp_config:
     description: "Additional MCP configuration (JSON string) that merges with the built-in GitHub MCP servers"
+  additional_permissions:
+    description: "Additional permissions to enable. Currently supports 'actions: read' for viewing workflow results"
+    required: false
+    default: ""
   claude_env:
     description: "Custom environment variables to pass to Claude Code execution (YAML format)"
     required: false
@@ -50,6 +65,9 @@ inputs:
   anthropic_api_key:
     description: "Anthropic API key (required for direct API, not needed for Bedrock/Vertex)"
     required: false
+  claude_code_oauth_token:
+    description: "Claude Code OAuth token (alternative to anthropic_api_key)"
+    required: false
   github_token:
     description: "GitHub token with repo and pull request permissions (optional if using GitHub App)"
     required: false
@@ -70,6 +88,10 @@ inputs:
     description: "Timeout in minutes for execution"
     required: false
     default: "30"
+  use_sticky_comment:
+    description: "Use just one comment to deliver issue/PR comments"
+    required: false
+    default: "false"
 
 outputs:
   execution_file:
@@ -98,7 +120,9 @@ runs:
       env:
         TRIGGER_PHRASE: ${{ inputs.trigger_phrase }}
         ASSIGNEE_TRIGGER: ${{ inputs.assignee_trigger }}
+        LABEL_TRIGGER: ${{ inputs.label_trigger }}
         BASE_BRANCH: ${{ inputs.base_branch }}
+        BRANCH_PREFIX: ${{ inputs.branch_prefix }}
         ALLOWED_TOOLS: ${{ inputs.allowed_tools }}
         DISALLOWED_TOOLS: ${{ inputs.disallowed_tools }}
         CUSTOM_INSTRUCTIONS: ${{ inputs.custom_instructions }}
@@ -106,11 +130,14 @@ runs:
         MCP_CONFIG: ${{ inputs.mcp_config }}
         OVERRIDE_GITHUB_TOKEN: ${{ inputs.github_token }}
         GITHUB_RUN_ID: ${{ github.run_id }}
+        USE_STICKY_COMMENT: ${{ inputs.use_sticky_comment }}
+        ACTIONS_TOKEN: ${{ github.token }}
+        ADDITIONAL_PERMISSIONS: ${{ inputs.additional_permissions }}
 
     - name: Run Claude Code
       id: claude-code
       if: steps.prepare.outputs.contains_trigger == 'true'
-      uses: anthropics/claude-code-base-action@bb2ef1d9768b9e94083d377778120f8f27958a72 # v0.0.22
+      uses: anthropics/claude-code-base-action@3560d21b41bd19b1d3ac6c9000af378903d8df0e # v0.0.32
       with:
         prompt_file: ${{ runner.temp }}/claude-prompts/claude-prompt.txt
         allowed_tools: ${{ env.ALLOWED_TOOLS }}
@@ -118,15 +145,18 @@ runs:
         timeout_minutes: ${{ inputs.timeout_minutes }}
         max_turns: ${{ inputs.max_turns }}
         model: ${{ inputs.model || inputs.anthropic_model }}
+        fallback_model: ${{ inputs.fallback_model }}
         mcp_config: ${{ steps.prepare.outputs.mcp_config }}
         use_bedrock: ${{ inputs.use_bedrock }}
         use_vertex: ${{ inputs.use_vertex }}
         anthropic_api_key: ${{ inputs.anthropic_api_key }}
+        claude_code_oauth_token: ${{ inputs.claude_code_oauth_token }}
         claude_env: ${{ inputs.claude_env }}
       env:
         # Model configuration
         ANTHROPIC_MODEL: ${{ inputs.model || inputs.anthropic_model }}
         GITHUB_TOKEN: ${{ steps.prepare.outputs.GITHUB_TOKEN }}
+        NODE_VERSION: ${{ env.NODE_VERSION }}
 
         # Provider configuration
         ANTHROPIC_BASE_URL: ${{ env.ANTHROPIC_BASE_URL }}
@@ -170,15 +200,24 @@ runs:
         TRIGGER_USERNAME: ${{ github.event.comment.user.login || github.event.issue.user.login || github.event.pull_request.user.login || github.event.sender.login || github.triggering_actor || github.actor || '' }}
         PREPARE_SUCCESS: ${{ steps.prepare.outcome == 'success' }}
         PREPARE_ERROR: ${{ steps.prepare.outputs.prepare_error || '' }}
+        USE_STICKY_COMMENT: ${{ inputs.use_sticky_comment }}
 
     - name: Display Claude Code Report
       if: steps.prepare.outputs.contains_trigger == 'true' && steps.claude-code.outputs.execution_file != ''
       shell: bash
       run: |
-        echo "## Claude Code Report" >> $GITHUB_STEP_SUMMARY
+        # Try to format the turns, but if it fails, dump the raw JSON
+        if bun run ${{ github.action_path }}/src/entrypoints/format-turns.ts "${{ steps.claude-code.outputs.execution_file }}" >> $GITHUB_STEP_SUMMARY 2>/dev/null; then
+          echo "Successfully formatted Claude Code report"
+        else
+          echo "## Claude Code Report (Raw Output)" >> $GITHUB_STEP_SUMMARY
+          echo "" >> $GITHUB_STEP_SUMMARY
+          echo "Failed to format output (please report). Here's the raw JSON:" >> $GITHUB_STEP_SUMMARY
+          echo "" >> $GITHUB_STEP_SUMMARY
           echo '```json' >> $GITHUB_STEP_SUMMARY
           cat "${{ steps.claude-code.outputs.execution_file }}" >> $GITHUB_STEP_SUMMARY
           echo '```' >> $GITHUB_STEP_SUMMARY
+        fi
 
     - name: Revoke app token
       if: always() && inputs.github_token == ''

examples/claude.yml

@@ -33,4 +33,6 @@ jobs:
         uses: anthropics/claude-code-action@beta
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          # Or use OAuth token instead:
+          # claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
           timeout_minutes: "60"

src/create-prompt/index.ts

@@ -36,9 +36,21 @@ const BASE_ALLOWED_TOOLS = [
 ];
 const DISALLOWED_TOOLS = ["WebSearch", "WebFetch"];
 
-export function buildAllowedToolsString(customAllowedTools?: string[]): string {
+export function buildAllowedToolsString(
+  customAllowedTools?: string[],
+  includeActionsTools: boolean = false,
+): string {
   let baseTools = [...BASE_ALLOWED_TOOLS];
 
+  // Add GitHub Actions MCP tools if enabled
+  if (includeActionsTools) {
+    baseTools.push(
+      "mcp__github_ci__get_ci_status",
+      "mcp__github_ci__get_workflow_run_details",
+      "mcp__github_ci__download_job_log",
+    );
+  }
+
   let allAllowedTools = baseTools.join(",");
   if (customAllowedTools && customAllowedTools.length > 0) {
     allAllowedTools = `${allAllowedTools},${customAllowedTools.join(",")}`;
@@ -81,6 +93,7 @@ export function prepareContext(
   const eventAction = context.eventAction;
   const triggerPhrase = context.inputs.triggerPhrase || "@claude";
   const assigneeTrigger = context.inputs.assigneeTrigger;
+  const labelTrigger = context.inputs.labelTrigger;
   const customInstructions = context.inputs.customInstructions;
   const allowedTools = context.inputs.allowedTools;
   const disallowedTools = context.inputs.disallowedTools;
@@ -242,7 +255,7 @@ export function prepareContext(
       }
 
       if (eventAction === "assigned") {
-        if (!assigneeTrigger) {
+        if (!assigneeTrigger && !directPrompt) {
           throw new Error(
             "ASSIGNEE_TRIGGER is required for issue assigned event",
           );
@@ -254,7 +267,20 @@ export function prepareContext(
           issueNumber,
           baseBranch,
           claudeBranch,
-          assigneeTrigger,
+          ...(assigneeTrigger && { assigneeTrigger }),
+        };
+      } else if (eventAction === "labeled") {
+        if (!labelTrigger) {
+          throw new Error("LABEL_TRIGGER is required for issue labeled event");
+        }
+        eventData = {
+          eventName: "issues",
+          eventAction: "labeled",
+          isPR: false,
+          issueNumber,
+          baseBranch,
+          claudeBranch,
+          labelTrigger,
         };
       } else if (eventAction === "opened") {
         eventData = {
@@ -328,10 +354,17 @@ export function getEventTypeAndContext(envVars: PreparedContext): {
           eventType: "ISSUE_CREATED",
           triggerContext: `new issue with '${envVars.triggerPhrase}' in body`,
         };
+      } else if (eventData.eventAction === "labeled") {
+        return {
+          eventType: "ISSUE_LABELED",
+          triggerContext: `issue labeled with '${eventData.labelTrigger}'`,
+        };
       }
       return {
         eventType: "ISSUE_ASSIGNED",
-        triggerContext: `issue assigned to '${eventData.assigneeTrigger}'`,
+        triggerContext: eventData.assigneeTrigger
+          ? `issue assigned to '${eventData.assigneeTrigger}'`
+          : `issue assigned event`,
       };
 
     case "pull_request":
@@ -465,6 +498,7 @@ Follow these steps:
    - Analyze the pre-fetched data provided above.
    - For ISSUE_CREATED: Read the issue body to find the request after the trigger phrase.
    - For ISSUE_ASSIGNED: Read the entire issue body to understand the task.
+   - For ISSUE_LABELED: Read the entire issue body to understand the task.
 ${eventData.eventName === "issue_comment" || eventData.eventName === "pull_request_review_comment" || eventData.eventName === "pull_request_review" ? `   - For comment/review events: Your instructions are in the <trigger_comment> tag above.` : ""}
 ${context.directPrompt ? `   - DIRECT INSTRUCTION: A direct instruction was provided and is shown in the <direct_prompt> tag above. This is not from any GitHub comment but a direct instruction to execute.` : ""}
    - IMPORTANT: Only the comment/issue containing '${context.triggerPhrase}' has your instructions.
@@ -643,8 +677,12 @@ export async function createPrompt(
     );
 
     // Set allowed tools
+    const hasActionsReadPermission =
+      context.inputs.additionalPermissions.get("actions") === "read" &&
+      context.isPR;
     const allAllowedTools = buildAllowedToolsString(
       context.inputs.allowedTools,
+      hasActionsReadPermission,
     );
     const allDisallowedTools = buildDisallowedToolsString(
       context.inputs.disallowedTools,

src/create-prompt/types.ts

@@ -65,7 +65,17 @@ type IssueAssignedEvent = {
   issueNumber: string;
   baseBranch: string;
   claudeBranch: string;
-  assigneeTrigger: string;
+  assigneeTrigger?: string;
+};
+
+type IssueLabeledEvent = {
+  eventName: "issues";
+  eventAction: "labeled";
+  isPR: false;
+  issueNumber: string;
+  baseBranch: string;
+  claudeBranch: string;
+  labelTrigger: string;
 };
 
 type PullRequestEvent = {
@@ -85,6 +95,7 @@ export type EventData =
   | IssueCommentEvent
   | IssueOpenedEvent
   | IssueAssignedEvent
+  | IssueLabeledEvent
   | PullRequestEvent;
 
 // Combined type with separate eventData field

src/entrypoints/format-turns.ts

@@ -0,0 +1,461 @@
+#!/usr/bin/env bun
+
+import { readFileSync, existsSync } from "fs";
+import { exit } from "process";
+
+export interface ToolUse {
+  type: string;
+  name?: string;
+  input?: Record<string, any>;
+  id?: string;
+}
+
+export interface ToolResult {
+  type: string;
+  tool_use_id?: string;
+  content?: any;
+  is_error?: boolean;
+}
+
+export interface ContentItem {
+  type: string;
+  text?: string;
+  tool_use_id?: string;
+  content?: any;
+  is_error?: boolean;
+  name?: string;
+  input?: Record<string, any>;
+  id?: string;
+}
+
+export interface Message {
+  content: ContentItem[];
+  usage?: {
+    input_tokens?: number;
+    output_tokens?: number;
+  };
+}
+
+export interface Turn {
+  type: string;
+  subtype?: string;
+  message?: Message;
+  tools?: any[];
+  cost_usd?: number;
+  duration_ms?: number;
+  result?: string;
+}
+
+export interface GroupedContent {
+  type: string;
+  tools_count?: number;
+  data?: Turn;
+  text_parts?: string[];
+  tool_calls?: { tool_use: ToolUse; tool_result?: ToolResult }[];
+  usage?: Record<string, number>;
+}
+
+export function detectContentType(content: any): string {
+  const contentStr = String(content).trim();
+
+  // Check for JSON
+  if (contentStr.startsWith("{") && contentStr.endsWith("}")) {
+    try {
+      JSON.parse(contentStr);
+      return "json";
+    } catch {
+      // Fall through
+    }
+  }
+
+  if (contentStr.startsWith("[") && contentStr.endsWith("]")) {
+    try {
+      JSON.parse(contentStr);
+      return "json";
+    } catch {
+      // Fall through
+    }
+  }
+
+  // Check for code-like content
+  const codeKeywords = [
+    "def ",
+    "class ",
+    "import ",
+    "from ",
+    "function ",
+    "const ",
+    "let ",
+    "var ",
+  ];
+  if (codeKeywords.some((keyword) => contentStr.includes(keyword))) {
+    if (
+      contentStr.includes("def ") ||
+      contentStr.includes("import ") ||
+      contentStr.includes("from ")
+    ) {
+      return "python";
+    } else if (
+      ["function ", "const ", "let ", "var ", "=>"].some((js) =>
+        contentStr.includes(js),
+      )
+    ) {
+      return "javascript";
+    } else {
+      return "python"; // default for code
+    }
+  }
+
+  // Check for shell/bash output
+  const shellIndicators = ["ls -", "cd ", "mkdir ", "rm ", "$ ", "# "];
+  if (
+    contentStr.startsWith("/") ||
+    contentStr.includes("Error:") ||
+    contentStr.startsWith("total ") ||
+    shellIndicators.some((indicator) => contentStr.includes(indicator))
+  ) {
+    return "bash";
+  }
+
+  // Check for diff format
+  if (
+    contentStr.startsWith("@@") ||
+    contentStr.includes("+++ ") ||
+    contentStr.includes("--- ")
+  ) {
+    return "diff";
+  }
+
+  // Check for HTML/XML
+  if (contentStr.startsWith("<") && contentStr.endsWith(">")) {
+    return "html";
+  }
+
+  // Check for markdown
+  const mdIndicators = ["# ", "## ", "### ", "- ", "* ", "```"];
+  if (mdIndicators.some((indicator) => contentStr.includes(indicator))) {
+    return "markdown";
+  }
+
+  // Default to plain text
+  return "text";
+}
+
+export function formatResultContent(content: any): string {
+  if (!content) {
+    return "*(No output)*\n\n";
+  }
+
+  let contentStr: string;
+
+  // Check if content is a list with "type": "text" structure
+  try {
+    let parsedContent: any;
+    if (typeof content === "string") {
+      parsedContent = JSON.parse(content);
+    } else {
+      parsedContent = content;
+    }
+
+    if (
+      Array.isArray(parsedContent) &&
+      parsedContent.length > 0 &&
+      typeof parsedContent[0] === "object" &&
+      parsedContent[0]?.type === "text"
+    ) {
+      // Extract the text field from the first item
+      contentStr = parsedContent[0]?.text || "";
+    } else {
+      contentStr = String(content).trim();
+    }
+  } catch {
+    contentStr = String(content).trim();
+  }
+
+  // Truncate very long results
+  if (contentStr.length > 3000) {
+    contentStr = contentStr.substring(0, 2997) + "...";
+  }
+
+  // Detect content type
+  const contentType = detectContentType(contentStr);
+
+  // Handle JSON content specially - pretty print it
+  if (contentType === "json") {
+    try {
+      // Try to parse and pretty print JSON
+      const parsed = JSON.parse(contentStr);
+      contentStr = JSON.stringify(parsed, null, 2);
+    } catch {
+      // Keep original if parsing fails
+    }
+  }
+
+  // Format with appropriate syntax highlighting
+  if (
+    contentType === "text" &&
+    contentStr.length < 100 &&
+    !contentStr.includes("\n")
+  ) {
+    // Short text results don't need code blocks
+    return `**→** ${contentStr}\n\n`;
+  } else {
+    return `**Result:**\n\`\`\`${contentType}\n${contentStr}\n\`\`\`\n\n`;
+  }
+}
+
+export function formatToolWithResult(
+  toolUse: ToolUse,
+  toolResult?: ToolResult,
+): string {
+  const toolName = toolUse.name || "unknown_tool";
+  const toolInput = toolUse.input || {};
+
+  let result = `### 🔧 \`${toolName}\`\n\n`;
+
+  // Add parameters if they exist and are not empty
+  if (Object.keys(toolInput).length > 0) {
+    result += "**Parameters:**\n```json\n";
+    result += JSON.stringify(toolInput, null, 2);
+    result += "\n```\n\n";
+  }
+
+  // Add result if available
+  if (toolResult) {
+    const content = toolResult.content || "";
+    const isError = toolResult.is_error || false;
+
+    if (isError) {
+      result += `❌ **Error:** \`${content}\`\n\n`;
+    } else {
+      result += formatResultContent(content);
+    }
+  }
+
+  return result;
+}
+
+export function groupTurnsNaturally(data: Turn[]): GroupedContent[] {
+  const groupedContent: GroupedContent[] = [];
+  const toolResultsMap = new Map<string, ToolResult>();
+
+  // First pass: collect all tool results by tool_use_id
+  for (const turn of data) {
+    if (turn.type === "user") {
+      const content = turn.message?.content || [];
+      for (const item of content) {
+        if (item.type === "tool_result" && item.tool_use_id) {
+          toolResultsMap.set(item.tool_use_id, {
+            type: item.type,
+            tool_use_id: item.tool_use_id,
+            content: item.content,
+            is_error: item.is_error,
+          });
+        }
+      }
+    }
+  }
+
+  // Second pass: process turns and group naturally
+  for (const turn of data) {
+    const turnType = turn.type || "unknown";
+
+    if (turnType === "system") {
+      const subtype = turn.subtype || "";
+      if (subtype === "init") {
+        const tools = turn.tools || [];
+        groupedContent.push({
+          type: "system_init",
+          tools_count: tools.length,
+        });
+      } else {
+        groupedContent.push({
+          type: "system_other",
+          data: turn,
+        });
+      }
+    } else if (turnType === "assistant") {
+      const message = turn.message || { content: [] };
+      const content = message.content || [];
+      const usage = message.usage || {};
+
+      // Process content items
+      const textParts: string[] = [];
+      const toolCalls: { tool_use: ToolUse; tool_result?: ToolResult }[] = [];
+
+      for (const item of content) {
+        const itemType = item.type || "";
+
+        if (itemType === "text") {
+          textParts.push(item.text || "");
+        } else if (itemType === "tool_use") {
+          const toolUseId = item.id;
+          const toolResult = toolUseId
+            ? toolResultsMap.get(toolUseId)
+            : undefined;
+          toolCalls.push({
+            tool_use: {
+              type: item.type,
+              name: item.name,
+              input: item.input,
+              id: item.id,
+            },
+            tool_result: toolResult,
+          });
+        }
+      }
+
+      if (textParts.length > 0 || toolCalls.length > 0) {
+        groupedContent.push({
+          type: "assistant_action",
+          text_parts: textParts,
+          tool_calls: toolCalls,
+          usage: usage,
+        });
+      }
+    } else if (turnType === "user") {
+      // Handle user messages that aren't tool results
+      const message = turn.message || { content: [] };
+      const content = message.content || [];
+      const textParts: string[] = [];
+
+      for (const item of content) {
+        if (item.type === "text") {
+          textParts.push(item.text || "");
+        }
+      }
+
+      if (textParts.length > 0) {
+        groupedContent.push({
+          type: "user_message",
+          text_parts: textParts,
+        });
+      }
+    } else if (turnType === "result") {
+      groupedContent.push({
+        type: "final_result",
+        data: turn,
+      });
+    }
+  }
+
+  return groupedContent;
+}
+
+export function formatGroupedContent(groupedContent: GroupedContent[]): string {
+  let markdown = "## Claude Code Report\n\n";
+
+  for (const item of groupedContent) {
+    const itemType = item.type;
+
+    if (itemType === "system_init") {
+      markdown += `## 🚀 System Initialization\n\n**Available Tools:** ${item.tools_count} tools loaded\n\n---\n\n`;
+    } else if (itemType === "system_other") {
+      markdown += `## ⚙️ System Message\n\n${JSON.stringify(item.data, null, 2)}\n\n---\n\n`;
+    } else if (itemType === "assistant_action") {
+      // Add text content first (if any) - no header needed
+      for (const text of item.text_parts || []) {
+        if (text.trim()) {
+          markdown += `${text}\n\n`;
+        }
+      }
+
+      // Add tool calls with their results
+      for (const toolCall of item.tool_calls || []) {
+        markdown += formatToolWithResult(
+          toolCall.tool_use,
+          toolCall.tool_result,
+        );
+      }
+
+      // Add usage info if available
+      const usage = item.usage || {};
+      if (Object.keys(usage).length > 0) {
+        const inputTokens = usage.input_tokens || 0;
+        const outputTokens = usage.output_tokens || 0;
+        markdown += `*Token usage: ${inputTokens} input, ${outputTokens} output*\n\n`;
+      }
+
+      // Only add separator if this section had content
+      if (
+        (item.text_parts && item.text_parts.length > 0) ||
+        (item.tool_calls && item.tool_calls.length > 0)
+      ) {
+        markdown += "---\n\n";
+      }
+    } else if (itemType === "user_message") {
+      markdown += "## 👤 User\n\n";
+      for (const text of item.text_parts || []) {
+        if (text.trim()) {
+          markdown += `${text}\n\n`;
+        }
+      }
+      markdown += "---\n\n";
+    } else if (itemType === "final_result") {
+      const data = item.data || {};
+      const cost = (data as any).cost_usd || 0;
+      const duration = (data as any).duration_ms || 0;
+      const resultText = (data as any).result || "";
+
+      markdown += "## ✅ Final Result\n\n";
+      if (resultText) {
+        markdown += `${resultText}\n\n`;
+      }
+      markdown += `**Cost:** $${cost.toFixed(4)} | **Duration:** ${(duration / 1000).toFixed(1)}s\n\n`;
+    }
+  }
+
+  return markdown;
+}
+
+export function formatTurnsFromData(data: Turn[]): string {
+  // Group turns naturally
+  const groupedContent = groupTurnsNaturally(data);
+
+  // Generate markdown
+  const markdown = formatGroupedContent(groupedContent);
+
+  return markdown;
+}
+
+function main(): void {
+  // Get the JSON file path from command line arguments
+  const args = process.argv.slice(2);
+  if (args.length === 0) {
+    console.error("Usage: format-turns.ts <json-file>");
+    exit(1);
+  }
+
+  const jsonFile = args[0];
+  if (!jsonFile) {
+    console.error("Error: No JSON file provided");
+    exit(1);
+  }
+
+  if (!existsSync(jsonFile)) {
+    console.error(`Error: ${jsonFile} not found`);
+    exit(1);
+  }
+
+  try {
+    // Read the JSON file
+    const fileContent = readFileSync(jsonFile, "utf-8");
+    const data: Turn[] = JSON.parse(fileContent);
+
+    // Group turns naturally
+    const groupedContent = groupTurnsNaturally(data);
+
+    // Generate markdown
+    const markdown = formatGroupedContent(groupedContent);
+
+    // Print to stdout (so it can be captured by shell)
+    console.log(markdown);
+  } catch (error) {
+    console.error(`Error processing file: ${error}`);
+    exit(1);
+  }
+}
+
+if (import.meta.main) {
+  main();
+}

src/entrypoints/prepare.ts

@@ -94,6 +94,7 @@ async function run() {
       additionalMcpConfig,
       claudeCommentId: commentId.toString(),
       allowedTools: context.inputs.allowedTools,
+      context,
     });
     core.setOutput("mcp_config", mcpConfig);
   } catch (error) {

src/github/context.ts

@@ -29,11 +29,15 @@ export type ParsedGitHubContext = {
   inputs: {
     triggerPhrase: string;
     assigneeTrigger: string;
+    labelTrigger: string;
     allowedTools: string[];
     disallowedTools: string[];
     customInstructions: string;
     directPrompt: string;
     baseBranch?: string;
+    branchPrefix: string;
+    useStickyComment: boolean;
+    additionalPermissions: Map<string, string>;
   };
 };
 
@@ -53,11 +57,17 @@ export function parseGitHubContext(): ParsedGitHubContext {
     inputs: {
       triggerPhrase: process.env.TRIGGER_PHRASE ?? "@claude",
       assigneeTrigger: process.env.ASSIGNEE_TRIGGER ?? "",
+      labelTrigger: process.env.LABEL_TRIGGER ?? "",
       allowedTools: parseMultilineInput(process.env.ALLOWED_TOOLS ?? ""),
       disallowedTools: parseMultilineInput(process.env.DISALLOWED_TOOLS ?? ""),
       customInstructions: process.env.CUSTOM_INSTRUCTIONS ?? "",
       directPrompt: process.env.DIRECT_PROMPT ?? "",
       baseBranch: process.env.BASE_BRANCH,
+      branchPrefix: process.env.BRANCH_PREFIX ?? "claude/",
+      useStickyComment: process.env.USE_STICKY_COMMENT === "true",
+      additionalPermissions: parseAdditionalPermissions(
+        process.env.ADDITIONAL_PERMISSIONS ?? "",
+      ),
     },
   };
 
@@ -119,6 +129,25 @@ export function parseMultilineInput(s: string): string[] {
     .filter((tool) => tool.length > 0);
 }
 
+export function parseAdditionalPermissions(s: string): Map<string, string> {
+  const permissions = new Map<string, string>();
+  if (!s || !s.trim()) {
+    return permissions;
+  }
+
+  const lines = s.trim().split("\n");
+  for (const line of lines) {
+    const trimmedLine = line.trim();
+    if (trimmedLine) {
+      const [key, value] = trimmedLine.split(":").map((part) => part.trim());
+      if (key && value) {
+        permissions.set(key, value);
+      }
+    }
+  }
+  return permissions;
+}
+
 export function isIssuesEvent(
   context: ParsedGitHubContext,
 ): context is ParsedGitHubContext & { payload: IssuesEvent } {

src/github/operations/branch.ts

@@ -26,7 +26,7 @@ export async function setupBranch(
 ): Promise<BranchInfo> {
   const { owner, repo } = context.repository;
   const entityNumber = context.entityNumber;
-  const { baseBranch } = context.inputs;
+  const { baseBranch, branchPrefix } = context.inputs;
   const isPR = context.isPR;
 
   if (isPR) {
@@ -97,7 +97,7 @@ export async function setupBranch(
     .split("T")
     .join("_");
 
-  const newBranch = `claude/${entityType}-${entityNumber}-${timestamp}`;
+  const newBranch = `${branchPrefix}${entityType}-${entityNumber}-${timestamp}`;
 
   try {
     // Get the SHA of the source branch

src/github/operations/comments/create-initial.ts

@@ -9,10 +9,13 @@ import { appendFileSync } from "fs";
 import { createJobRunLink, createCommentBody } from "./common";
 import {
   isPullRequestReviewCommentEvent,
+  isPullRequestEvent,
   type ParsedGitHubContext,
 } from "../../context";
 import type { Octokit } from "@octokit/rest";
 
+const CLAUDE_APP_BOT_ID = 209825114;
+
 export async function createInitialComment(
   octokit: Octokit,
   context: ParsedGitHubContext,
@@ -25,8 +28,43 @@ export async function createInitialComment(
   try {
     let response;
 
+    if (
+      context.inputs.useStickyComment &&
+      context.isPR &&
+      isPullRequestEvent(context)
+    ) {
+      const comments = await octokit.rest.issues.listComments({
+        owner,
+        repo,
+        issue_number: context.entityNumber,
+      });
+      const existingComment = comments.data.find((comment) => {
+        const idMatch = comment.user?.id === CLAUDE_APP_BOT_ID;
+        const botNameMatch =
+          comment.user?.type === "Bot" &&
+          comment.user?.login.toLowerCase().includes("claude");
+        const bodyMatch = comment.body === initialBody;
+
+        return idMatch || botNameMatch || bodyMatch;
+      });
+      if (existingComment) {
+        response = await octokit.rest.issues.updateComment({
+          owner,
+          repo,
+          comment_id: existingComment.id,
+          body: initialBody,
+        });
+      } else {
+        // Create new comment if no existing one found
+        response = await octokit.rest.issues.createComment({
+          owner,
+          repo,
+          issue_number: context.entityNumber,
+          body: initialBody,
+        });
+      }
+    } else if (isPullRequestReviewCommentEvent(context)) {
       // Only use createReplyForReviewComment if it's a PR review comment AND we have a comment_id
-    if (isPullRequestReviewCommentEvent(context)) {
       response = await octokit.rest.pulls.createReplyForReviewComment({
         owner,
         repo,

src/github/token.ts

@@ -1,47 +1,7 @@
 #!/usr/bin/env bun
 
 import * as core from "@actions/core";
-
+import { retryWithBackoff } from "../utils/retry";
-type RetryOptions = {
-  maxAttempts?: number;
-  initialDelayMs?: number;
-  maxDelayMs?: number;
-  backoffFactor?: number;
-};
-
-async function retryWithBackoff<T>(
-  operation: () => Promise<T>,
-  options: RetryOptions = {},
-): Promise<T> {
-  const {
-    maxAttempts = 3,
-    initialDelayMs = 5000,
-    maxDelayMs = 20000,
-    backoffFactor = 2,
-  } = options;
-
-  let delayMs = initialDelayMs;
-  let lastError: Error | undefined;
-
-  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
-    try {
-      console.log(`Attempt ${attempt} of ${maxAttempts}...`);
-      return await operation();
-    } catch (error) {
-      lastError = error instanceof Error ? error : new Error(String(error));
-      console.error(`Attempt ${attempt} failed:`, lastError.message);
-
-      if (attempt < maxAttempts) {
-        console.log(`Retrying in ${delayMs / 1000} seconds...`);
-        await new Promise((resolve) => setTimeout(resolve, delayMs));
-        delayMs = Math.min(delayMs * backoffFactor, maxDelayMs);
-      }
-    }
-  }
-
-  console.error(`Operation failed after ${maxAttempts} attempts`);
-  throw lastError;
-}
 
 async function getOidcToken(): Promise<string> {
   try {

src/github/validation/trigger.ts

@@ -13,7 +13,7 @@ import type { ParsedGitHubContext } from "../context";
 
 export function checkContainsTrigger(context: ParsedGitHubContext): boolean {
   const {
-    inputs: { assigneeTrigger, triggerPhrase, directPrompt },
+    inputs: { assigneeTrigger, labelTrigger, triggerPhrase, directPrompt },
   } = context;
 
   // If direct prompt is provided, always trigger
@@ -34,6 +34,16 @@ export function checkContainsTrigger(context: ParsedGitHubContext): boolean {
     }
   }
 
+  // Check for label trigger
+  if (isIssuesEvent(context) && context.eventAction === "labeled") {
+    const labelName = (context.payload as any).label?.name || "";
+
+    if (labelTrigger && labelName === labelTrigger) {
+      console.log(`Issue labeled with trigger label '${labelTrigger}'`);
+      return true;
+    }
+  }
+
   // Check for issue body and title trigger on issue creation
   if (isIssuesEvent(context) && context.eventAction === "opened") {
     const issueBody = context.payload.issue.body || "";

src/mcp/github-actions-server.ts

@@ -0,0 +1,275 @@
+#!/usr/bin/env node
+
+import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { z } from "zod";
+import { mkdir, writeFile } from "fs/promises";
+import { Octokit } from "@octokit/rest";
+
+const REPO_OWNER = process.env.REPO_OWNER;
+const REPO_NAME = process.env.REPO_NAME;
+const PR_NUMBER = process.env.PR_NUMBER;
+const GITHUB_TOKEN = process.env.GITHUB_TOKEN;
+const RUNNER_TEMP = process.env.RUNNER_TEMP || "/tmp";
+
+if (!REPO_OWNER || !REPO_NAME || !PR_NUMBER || !GITHUB_TOKEN) {
+  console.error(
+    "[GitHub CI Server] Error: REPO_OWNER, REPO_NAME, PR_NUMBER, and GITHUB_TOKEN environment variables are required",
+  );
+  process.exit(1);
+}
+
+const server = new McpServer({
+  name: "GitHub CI Server",
+  version: "0.0.1",
+});
+
+console.error("[GitHub CI Server] MCP Server instance created");
+
+server.tool(
+  "get_ci_status",
+  "Get CI status summary for this PR",
+  {
+    status: z
+      .enum([
+        "completed",
+        "action_required",
+        "cancelled",
+        "failure",
+        "neutral",
+        "skipped",
+        "stale",
+        "success",
+        "timed_out",
+        "in_progress",
+        "queued",
+        "requested",
+        "waiting",
+        "pending",
+      ])
+      .optional()
+      .describe("Filter workflow runs by status"),
+  },
+  async ({ status }) => {
+    try {
+      const client = new Octokit({
+        auth: GITHUB_TOKEN,
+      });
+
+      // Get the PR to find the head SHA
+      const { data: prData } = await client.pulls.get({
+        owner: REPO_OWNER!,
+        repo: REPO_NAME!,
+        pull_number: parseInt(PR_NUMBER!, 10),
+      });
+      const headSha = prData.head.sha;
+
+      const { data: runsData } = await client.actions.listWorkflowRunsForRepo({
+        owner: REPO_OWNER!,
+        repo: REPO_NAME!,
+        head_sha: headSha,
+        ...(status && { status }),
+      });
+
+      // Process runs to create summary
+      const runs = runsData.workflow_runs || [];
+      const summary = {
+        total_runs: runs.length,
+        failed: 0,
+        passed: 0,
+        pending: 0,
+      };
+
+      const processedRuns = runs.map((run: any) => {
+        // Update summary counts
+        if (run.status === "completed") {
+          if (run.conclusion === "success") {
+            summary.passed++;
+          } else if (run.conclusion === "failure") {
+            summary.failed++;
+          }
+        } else {
+          summary.pending++;
+        }
+
+        return {
+          id: run.id,
+          name: run.name,
+          status: run.status,
+          conclusion: run.conclusion,
+          html_url: run.html_url,
+          created_at: run.created_at,
+        };
+      });
+
+      const result = {
+        summary,
+        runs: processedRuns,
+      };
+
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(result, null, 2),
+          },
+        ],
+      };
+    } catch (error) {
+      const errorMessage =
+        error instanceof Error ? error.message : String(error);
+      return {
+        content: [
+          {
+            type: "text",
+            text: `Error: ${errorMessage}`,
+          },
+        ],
+        error: errorMessage,
+        isError: true,
+      };
+    }
+  },
+);
+
+server.tool(
+  "get_workflow_run_details",
+  "Get job and step details for a workflow run",
+  {
+    run_id: z.number().describe("The workflow run ID"),
+  },
+  async ({ run_id }) => {
+    try {
+      const client = new Octokit({
+        auth: GITHUB_TOKEN,
+      });
+
+      // Get jobs for this workflow run
+      const { data: jobsData } = await client.actions.listJobsForWorkflowRun({
+        owner: REPO_OWNER!,
+        repo: REPO_NAME!,
+        run_id,
+      });
+
+      const processedJobs = jobsData.jobs.map((job: any) => {
+        // Extract failed steps
+        const failedSteps = (job.steps || [])
+          .filter((step: any) => step.conclusion === "failure")
+          .map((step: any) => ({
+            name: step.name,
+            number: step.number,
+          }));
+
+        return {
+          id: job.id,
+          name: job.name,
+          conclusion: job.conclusion,
+          html_url: job.html_url,
+          failed_steps: failedSteps,
+        };
+      });
+
+      const result = {
+        jobs: processedJobs,
+      };
+
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(result, null, 2),
+          },
+        ],
+      };
+    } catch (error) {
+      const errorMessage =
+        error instanceof Error ? error.message : String(error);
+
+      return {
+        content: [
+          {
+            type: "text",
+            text: `Error: ${errorMessage}`,
+          },
+        ],
+        error: errorMessage,
+        isError: true,
+      };
+    }
+  },
+);
+
+server.tool(
+  "download_job_log",
+  "Download job logs to disk",
+  {
+    job_id: z.number().describe("The job ID"),
+  },
+  async ({ job_id }) => {
+    try {
+      const client = new Octokit({
+        auth: GITHUB_TOKEN,
+      });
+
+      const response = await client.actions.downloadJobLogsForWorkflowRun({
+        owner: REPO_OWNER!,
+        repo: REPO_NAME!,
+        job_id,
+      });
+
+      const logsText = response.data as unknown as string;
+
+      const logsDir = `${RUNNER_TEMP}/github-ci-logs`;
+      await mkdir(logsDir, { recursive: true });
+
+      const logPath = `${logsDir}/job-${job_id}.log`;
+      await writeFile(logPath, logsText, "utf-8");
+
+      const result = {
+        path: logPath,
+        size_bytes: Buffer.byteLength(logsText, "utf-8"),
+      };
+
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(result, null, 2),
+          },
+        ],
+      };
+    } catch (error) {
+      const errorMessage =
+        error instanceof Error ? error.message : String(error);
+
+      return {
+        content: [
+          {
+            type: "text",
+            text: `Error: ${errorMessage}`,
+          },
+        ],
+        error: errorMessage,
+        isError: true,
+      };
+    }
+  },
+);
+
+async function runServer() {
+  try {
+    const transport = new StdioServerTransport();
+
+    await server.connect(transport);
+
+    process.on("exit", () => {
+      server.close();
+    });
+  } catch (error) {
+    throw error;
+  }
+}
+
+runServer().catch(() => {
+  process.exit(1);
+});

src/mcp/github-file-ops-server.ts

@@ -9,6 +9,7 @@ import fetch from "node-fetch";
 import { GITHUB_API_URL } from "../github/api/config";
 import { Octokit } from "@octokit/rest";
 import { updateClaudeComment } from "../github/operations/comments/update-claude-comment";
+import { retryWithBackoff } from "../utils/retry";
 
 type GitHubRef = {
   object: {
@@ -125,6 +126,50 @@ server.tool(
             ? filePath
             : join(REPO_DIR, filePath);
 
+          // Check if file is binary (images, etc.)
+          const isBinaryFile =
+            /\.(png|jpg|jpeg|gif|webp|ico|pdf|zip|tar|gz|exe|bin|woff|woff2|ttf|eot)$/i.test(
+              filePath,
+            );
+
+          if (isBinaryFile) {
+            // For binary files, create a blob first using the Blobs API
+            const binaryContent = await readFile(fullPath);
+
+            // Create blob using Blobs API (supports encoding parameter)
+            const blobUrl = `${GITHUB_API_URL}/repos/${owner}/${repo}/git/blobs`;
+            const blobResponse = await fetch(blobUrl, {
+              method: "POST",
+              headers: {
+                Accept: "application/vnd.github+json",
+                Authorization: `Bearer ${githubToken}`,
+                "X-GitHub-Api-Version": "2022-11-28",
+                "Content-Type": "application/json",
+              },
+              body: JSON.stringify({
+                content: binaryContent.toString("base64"),
+                encoding: "base64",
+              }),
+            });
+
+            if (!blobResponse.ok) {
+              const errorText = await blobResponse.text();
+              throw new Error(
+                `Failed to create blob for ${filePath}: ${blobResponse.status} - ${errorText}`,
+              );
+            }
+
+            const blobData = (await blobResponse.json()) as { sha: string };
+
+            // Return tree entry with blob SHA
+            return {
+              path: filePath,
+              mode: "100644",
+              type: "blob",
+              sha: blobData.sha,
+            };
+          } else {
+            // For text files, include content directly in tree
             const content = await readFile(fullPath, "utf-8");
             return {
               path: filePath,
@@ -132,6 +177,7 @@ server.tool(
               type: "blob",
               content: content,
             };
+          }
         }),
       );
 
@@ -188,6 +234,12 @@ server.tool(
 
       // 6. Update the reference to point to the new commit
       const updateRefUrl = `${GITHUB_API_URL}/repos/${owner}/${repo}/git/refs/heads/${branch}`;
+
+      // We're seeing intermittent 403 "Resource not accessible by integration" errors
+      // on certain repos when updating git references. These appear to be transient
+      // GitHub API issues that succeed on retry.
+      await retryWithBackoff(
+        async () => {
           const updateRefResponse = await fetch(updateRefUrl, {
             method: "PATCH",
             headers: {
@@ -204,11 +256,29 @@ server.tool(
 
           if (!updateRefResponse.ok) {
             const errorText = await updateRefResponse.text();
-        throw new Error(
+            const error = new Error(
               `Failed to update reference: ${updateRefResponse.status} - ${errorText}`,
             );
+
+            // Only retry on 403 errors - these are the intermittent failures we're targeting
+            if (updateRefResponse.status === 403) {
+              console.log("Received 403 error, will retry...");
+              throw error;
             }
 
+            // For non-403 errors, fail immediately without retry
+            console.error("Non-retryable error:", updateRefResponse.status);
+            throw error;
+          }
+        },
+        {
+          maxAttempts: 3,
+          initialDelayMs: 1000, // Start with 1 second delay
+          maxDelayMs: 5000, // Max 5 seconds delay
+          backoffFactor: 2, // Double the delay each time
+        },
+      );
+
       const simplifiedResult = {
         commit: {
           sha: newCommitData.sha,
@@ -382,6 +452,12 @@ server.tool(
 
       // 6. Update the reference to point to the new commit
       const updateRefUrl = `${GITHUB_API_URL}/repos/${owner}/${repo}/git/refs/heads/${branch}`;
+
+      // We're seeing intermittent 403 "Resource not accessible by integration" errors
+      // on certain repos when updating git references. These appear to be transient
+      // GitHub API issues that succeed on retry.
+      await retryWithBackoff(
+        async () => {
           const updateRefResponse = await fetch(updateRefUrl, {
             method: "PATCH",
             headers: {
@@ -398,11 +474,29 @@ server.tool(
 
           if (!updateRefResponse.ok) {
             const errorText = await updateRefResponse.text();
-        throw new Error(
+            const error = new Error(
               `Failed to update reference: ${updateRefResponse.status} - ${errorText}`,
             );
+
+            // Only retry on 403 errors - these are the intermittent failures we're targeting
+            if (updateRefResponse.status === 403) {
+              console.log("Received 403 error, will retry...");
+              throw error;
             }
 
+            // For non-403 errors, fail immediately without retry
+            console.error("Non-retryable error:", updateRefResponse.status);
+            throw error;
+          }
+        },
+        {
+          maxAttempts: 3,
+          initialDelayMs: 1000, // Start with 1 second delay
+          maxDelayMs: 5000, // Max 5 seconds delay
+          backoffFactor: 2, // Double the delay each time
+        },
+      );
+
       const simplifiedResult = {
         commit: {
           sha: newCommitData.sha,

src/mcp/install-mcp-server.ts

@@ -1,5 +1,7 @@
 import * as core from "@actions/core";
 import { GITHUB_API_URL } from "../github/api/config";
+import type { ParsedGitHubContext } from "../github/context";
+import { Octokit } from "@octokit/rest";
 
 type PrepareConfigParams = {
   githubToken: string;
@@ -9,8 +11,41 @@ type PrepareConfigParams = {
   additionalMcpConfig?: string;
   claudeCommentId?: string;
   allowedTools: string[];
+  context: ParsedGitHubContext;
 };
 
+async function checkActionsReadPermission(
+  token: string,
+  owner: string,
+  repo: string,
+): Promise<boolean> {
+  try {
+    const client = new Octokit({ auth: token });
+
+    // Try to list workflow runs - this requires actions:read
+    // We use per_page=1 to minimize the response size
+    await client.actions.listWorkflowRunsForRepo({
+      owner,
+      repo,
+      per_page: 1,
+    });
+
+    return true;
+  } catch (error: any) {
+    // Check if it's a permission error
+    if (
+      error.status === 403 &&
+      error.message?.includes("Resource not accessible")
+    ) {
+      return false;
+    }
+
+    // For other errors (network issues, etc), log but don't fail
+    core.debug(`Failed to check actions permission: ${error.message}`);
+    return false;
+  }
+}
+
 export async function prepareMcpConfig(
   params: PrepareConfigParams,
 ): Promise<string> {
@@ -22,6 +57,7 @@ export async function prepareMcpConfig(
     additionalMcpConfig,
     claudeCommentId,
     allowedTools,
+    context,
   } = params;
   try {
     const allowedToolsList = allowedTools || [];
@@ -53,6 +89,42 @@ export async function prepareMcpConfig(
       },
     };
 
+    // Only add CI server if we have actions:read permission and we're in a PR context
+    const hasActionsReadPermission =
+      context.inputs.additionalPermissions.get("actions") === "read";
+
+    if (context.isPR && hasActionsReadPermission) {
+      // Verify the token actually has actions:read permission
+      const actuallyHasPermission = await checkActionsReadPermission(
+        process.env.ACTIONS_TOKEN || "",
+        owner,
+        repo,
+      );
+
+      if (!actuallyHasPermission) {
+        core.warning(
+          "The github_ci MCP server requires 'actions: read' permission. " +
+            "Please ensure your GitHub token has this permission. " +
+            "See: https://docs.github.com/en/actions/security-guides/automatic-token-authentication#permissions-for-the-github_token",
+        );
+      }
+      baseMcpConfig.mcpServers.github_ci = {
+        command: "bun",
+        args: [
+          "run",
+          `${process.env.GITHUB_ACTION_PATH}/src/mcp/github-actions-server.ts`,
+        ],
+        env: {
+          // Use workflow github token, not app token
+          GITHUB_TOKEN: process.env.ACTIONS_TOKEN,
+          REPO_OWNER: owner,
+          REPO_NAME: repo,
+          PR_NUMBER: context.entityNumber.toString(),
+          RUNNER_TEMP: process.env.RUNNER_TEMP || "/tmp",
+        },
+      };
+    }
+
     if (hasGitHubMcpTools) {
       baseMcpConfig.mcpServers.github = {
         command: "docker",
@@ -62,7 +134,7 @@ export async function prepareMcpConfig(
           "--rm",
           "-e",
           "GITHUB_PERSONAL_ACCESS_TOKEN",
-          "ghcr.io/github/github-mcp-server:sha-6d69797", // https://github.com/github/github-mcp-server/releases/tag/v0.5.0
+          "ghcr.io/github/github-mcp-server:sha-721fd3e", // https://github.com/github/github-mcp-server/releases/tag/v0.6.0
         ],
         env: {
           GITHUB_PERSONAL_ACCESS_TOKEN: githubToken,

src/utils/retry.ts

@@ -0,0 +1,40 @@
+export type RetryOptions = {
+  maxAttempts?: number;
+  initialDelayMs?: number;
+  maxDelayMs?: number;
+  backoffFactor?: number;
+};
+
+export async function retryWithBackoff<T>(
+  operation: () => Promise<T>,
+  options: RetryOptions = {},
+): Promise<T> {
+  const {
+    maxAttempts = 3,
+    initialDelayMs = 5000,
+    maxDelayMs = 20000,
+    backoffFactor = 2,
+  } = options;
+
+  let delayMs = initialDelayMs;
+  let lastError: Error | undefined;
+
+  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
+    try {
+      console.log(`Attempt ${attempt} of ${maxAttempts}...`);
+      return await operation();
+    } catch (error) {
+      lastError = error instanceof Error ? error : new Error(String(error));
+      console.error(`Attempt ${attempt} failed:`, lastError.message);
+
+      if (attempt < maxAttempts) {
+        console.log(`Retrying in ${delayMs / 1000} seconds...`);
+        await new Promise((resolve) => setTimeout(resolve, delayMs));
+        delayMs = Math.min(delayMs * backoffFactor, maxDelayMs);
+      }
+    }
+  }
+
+  console.error(`Operation failed after ${maxAttempts} attempts`);
+  throw lastError;
+}

test/create-prompt.test.ts

@@ -226,6 +226,33 @@ describe("generatePrompt", () => {
     );
   });
 
+  test("should generate prompt for issue labeled event", () => {
+    const envVars: PreparedContext = {
+      repository: "owner/repo",
+      claudeCommentId: "12345",
+      triggerPhrase: "@claude",
+      eventData: {
+        eventName: "issues",
+        eventAction: "labeled",
+        isPR: false,
+        issueNumber: "888",
+        baseBranch: "main",
+        claudeBranch: "claude/issue-888-20240101_120000",
+        labelTrigger: "claude-task",
+      },
+    };
+
+    const prompt = generatePrompt(envVars, mockGitHubData);
+
+    expect(prompt).toContain("<event_type>ISSUE_LABELED</event_type>");
+    expect(prompt).toContain(
+      "<trigger_context>issue labeled with 'claude-task'</trigger_context>",
+    );
+    expect(prompt).toContain(
+      "[Create a PR](https://github.com/owner/repo/compare/main",
+    );
+  });
+
   test("should include direct prompt when provided", () => {
     const envVars: PreparedContext = {
       repository: "owner/repo",
@@ -614,6 +641,51 @@ describe("getEventTypeAndContext", () => {
     expect(result.eventType).toBe("ISSUE_ASSIGNED");
     expect(result.triggerContext).toBe("issue assigned to 'claude-bot'");
   });
+
+  test("should return correct type and context for issue labeled", () => {
+    const envVars: PreparedContext = {
+      repository: "owner/repo",
+      claudeCommentId: "12345",
+      triggerPhrase: "@claude",
+      eventData: {
+        eventName: "issues",
+        eventAction: "labeled",
+        isPR: false,
+        issueNumber: "888",
+        baseBranch: "main",
+        claudeBranch: "claude/issue-888-20240101_120000",
+        labelTrigger: "claude-task",
+      },
+    };
+
+    const result = getEventTypeAndContext(envVars);
+
+    expect(result.eventType).toBe("ISSUE_LABELED");
+    expect(result.triggerContext).toBe("issue labeled with 'claude-task'");
+  });
+
+  test("should return correct type and context for issue assigned without assigneeTrigger", () => {
+    const envVars: PreparedContext = {
+      repository: "owner/repo",
+      claudeCommentId: "12345",
+      triggerPhrase: "@claude",
+      directPrompt: "Please assess this issue",
+      eventData: {
+        eventName: "issues",
+        eventAction: "assigned",
+        isPR: false,
+        issueNumber: "999",
+        baseBranch: "main",
+        claudeBranch: "claude/issue-999-20240101_120000",
+        // No assigneeTrigger when using directPrompt
+      },
+    };
+
+    const result = getEventTypeAndContext(envVars);
+
+    expect(result.eventType).toBe("ISSUE_ASSIGNED");
+    expect(result.triggerContext).toBe("issue assigned event");
+  });
 });
 
 describe("buildAllowedToolsString", () => {
@@ -671,6 +743,36 @@ describe("buildAllowedToolsString", () => {
     expect(basePlusCustom).toContain("Tool2");
     expect(basePlusCustom).toContain("Tool3");
   });
+
+  test("should include GitHub Actions tools when includeActionsTools is true", () => {
+    const result = buildAllowedToolsString([], true);
+
+    // Base tools should be present
+    expect(result).toContain("Edit");
+    expect(result).toContain("Glob");
+
+    // GitHub Actions tools should be included
+    expect(result).toContain("mcp__github_ci__get_ci_status");
+    expect(result).toContain("mcp__github_ci__get_workflow_run_details");
+    expect(result).toContain("mcp__github_ci__download_job_log");
+  });
+
+  test("should include both custom and Actions tools when both provided", () => {
+    const customTools = ["Tool1", "Tool2"];
+    const result = buildAllowedToolsString(customTools, true);
+
+    // Base tools should be present
+    expect(result).toContain("Edit");
+
+    // Custom tools should be included
+    expect(result).toContain("Tool1");
+    expect(result).toContain("Tool2");
+
+    // GitHub Actions tools should be included
+    expect(result).toContain("mcp__github_ci__get_ci_status");
+    expect(result).toContain("mcp__github_ci__get_workflow_run_details");
+    expect(result).toContain("mcp__github_ci__download_job_log");
+  });
 });
 
 describe("buildDisallowedToolsString", () => {

test/fixtures/sample-turns-expected-output.md

@@ -0,0 +1,95 @@
+## Claude Code Report
+
+## 🚀 System Initialization
+
+**Available Tools:** 8 tools loaded
+
+---
+
+I'll help you with this task. Let me start by examining the file to understand what needs to be changed.
+
+### 🔧 `Read`
+
+**Parameters:**
+```json
+{
+  "file_path": "/path/to/sample/file.py"
+}
+```
+
+**Result:**
+```python
+def example_function():
+    print("Debug message")  # This should be removed
+    return "Hello World"
+
+if __name__ == "__main__":
+    result = example_function()
+    print(result)
+```
+
+*Token usage: 100 input, 75 output*
+
+---
+
+I can see the debug print statement that needs to be removed. Let me fix this by editing the file.
+
+### 🔧 `Edit`
+
+**Parameters:**
+```json
+{
+  "file_path": "/path/to/sample/file.py",
+  "old_string": "def example_function():\n    print(\"Debug message\")  # This should be removed\n    return \"Hello World\"",
+  "new_string": "def example_function():\n    return \"Hello World\""
+}
+```
+
+**→** File successfully edited. The debug print statement has been removed.
+
+*Token usage: 200 input, 50 output*
+
+---
+
+Perfect! I've successfully removed the debug print statement from the function. Let me now create a pull request comment to document this change.
+
+### 🔧 `mcp__github__add_pull_request_review_comment`
+
+**Parameters:**
+```json
+{
+  "owner": "example-org",
+  "repo": "example-repo",
+  "pull_number": 123,
+  "body": "Removed debug print statement as requested.",
+  "commit_id": "abc123def456",
+  "path": "sample/file.py",
+  "line": 2
+}
+```
+
+**→** Successfully posted review comment to PR #123
+
+*Token usage: 150 input, 80 output*
+
+---
+
+Great! I've successfully completed the requested task:
+
+1. ✅ Located the debug print statement in the file
+2. ✅ Removed the print statement while preserving the function logic
+3. ✅ Added a review comment documenting the change
+
+The debug print statement has been removed as requested by the reviewers.
+
+*Token usage: 180 input, 60 output*
+
+---
+
+## ✅ Final Result
+
+Successfully removed debug print statement from file and added review comment to document the change.
+
+**Cost:** $0.0347 | **Duration:** 18.8s
+
+

test/fixtures/sample-turns.json

@@ -0,0 +1,196 @@
+[
+  {
+    "type": "system",
+    "subtype": "init",
+    "session_id": "sample-session-id",
+    "tools": [
+      "Task",
+      "Bash",
+      "Read",
+      "Edit",
+      "Write",
+      "mcp__github__get_file_contents",
+      "mcp__github__create_or_update_file",
+      "mcp__github__add_pull_request_review_comment"
+    ],
+    "mcp_servers": [
+      {
+        "name": "github",
+        "status": "connected"
+      }
+    ]
+  },
+  {
+    "type": "assistant",
+    "message": {
+      "id": "msg_sample123",
+      "type": "message",
+      "role": "assistant",
+      "model": "claude-test-model",
+      "content": [
+        {
+          "type": "text",
+          "text": "I'll help you with this task. Let me start by examining the file to understand what needs to be changed."
+        },
+        {
+          "type": "tool_use",
+          "id": "tool_call_1",
+          "name": "Read",
+          "input": {
+            "file_path": "/path/to/sample/file.py"
+          }
+        }
+      ],
+      "stop_reason": "tool_use",
+      "stop_sequence": null,
+      "usage": {
+        "input_tokens": 100,
+        "cache_creation_input_tokens": 0,
+        "cache_read_input_tokens": 50,
+        "output_tokens": 75
+      }
+    },
+    "session_id": "sample-session-id"
+  },
+  {
+    "type": "user",
+    "message": {
+      "content": [
+        {
+          "type": "tool_result",
+          "tool_use_id": "tool_call_1",
+          "content": "def example_function():\n    print(\"Debug message\")  # This should be removed\n    return \"Hello World\"\n\nif __name__ == \"__main__\":\n    result = example_function()\n    print(result)",
+          "is_error": false
+        }
+      ]
+    }
+  },
+  {
+    "type": "assistant",
+    "message": {
+      "id": "msg_sample124",
+      "type": "message",
+      "role": "assistant",
+      "model": "claude-test-model",
+      "content": [
+        {
+          "type": "text",
+          "text": "I can see the debug print statement that needs to be removed. Let me fix this by editing the file."
+        },
+        {
+          "type": "tool_use",
+          "id": "tool_call_2",
+          "name": "Edit",
+          "input": {
+            "file_path": "/path/to/sample/file.py",
+            "old_string": "def example_function():\n    print(\"Debug message\")  # This should be removed\n    return \"Hello World\"",
+            "new_string": "def example_function():\n    return \"Hello World\""
+          }
+        }
+      ],
+      "stop_reason": "tool_use",
+      "stop_sequence": null,
+      "usage": {
+        "input_tokens": 200,
+        "cache_creation_input_tokens": 0,
+        "cache_read_input_tokens": 100,
+        "output_tokens": 50
+      }
+    },
+    "session_id": "sample-session-id"
+  },
+  {
+    "type": "user",
+    "message": {
+      "content": [
+        {
+          "type": "tool_result",
+          "tool_use_id": "tool_call_2",
+          "content": "File successfully edited. The debug print statement has been removed.",
+          "is_error": false
+        }
+      ]
+    }
+  },
+  {
+    "type": "assistant",
+    "message": {
+      "id": "msg_sample125",
+      "type": "message",
+      "role": "assistant",
+      "model": "claude-test-model",
+      "content": [
+        {
+          "type": "text",
+          "text": "Perfect! I've successfully removed the debug print statement from the function. Let me now create a pull request comment to document this change."
+        },
+        {
+          "type": "tool_use",
+          "id": "tool_call_3",
+          "name": "mcp__github__add_pull_request_review_comment",
+          "input": {
+            "owner": "example-org",
+            "repo": "example-repo",
+            "pull_number": 123,
+            "body": "Removed debug print statement as requested.",
+            "commit_id": "abc123def456",
+            "path": "sample/file.py",
+            "line": 2
+          }
+        }
+      ],
+      "stop_reason": "tool_use",
+      "stop_sequence": null,
+      "usage": {
+        "input_tokens": 150,
+        "cache_creation_input_tokens": 0,
+        "cache_read_input_tokens": 75,
+        "output_tokens": 80
+      }
+    },
+    "session_id": "sample-session-id"
+  },
+  {
+    "type": "user",
+    "message": {
+      "content": [
+        {
+          "type": "tool_result",
+          "tool_use_id": "tool_call_3",
+          "content": "Successfully posted review comment to PR #123",
+          "is_error": false
+        }
+      ]
+    }
+  },
+  {
+    "type": "assistant",
+    "message": {
+      "id": "msg_sample126",
+      "type": "message",
+      "role": "assistant",
+      "model": "claude-test-model",
+      "content": [
+        {
+          "type": "text",
+          "text": "Great! I've successfully completed the requested task:\n\n1. ✅ Located the debug print statement in the file\n2. ✅ Removed the print statement while preserving the function logic\n3. ✅ Added a review comment documenting the change\n\nThe debug print statement has been removed as requested by the reviewers."
+        }
+      ],
+      "stop_reason": "end_turn",
+      "stop_sequence": null,
+      "usage": {
+        "input_tokens": 180,
+        "cache_creation_input_tokens": 0,
+        "cache_read_input_tokens": 90,
+        "output_tokens": 60
+      }
+    },
+    "session_id": "sample-session-id"
+  },
+  {
+    "type": "result",
+    "cost_usd": 0.0347,
+    "duration_ms": 18750,
+    "result": "Successfully removed debug print statement from file and added review comment to document the change."
+  }
+]

test/format-turns.test.ts

@@ -0,0 +1,439 @@
+import { expect, test, describe } from "bun:test";
+import { readFileSync } from "fs";
+import { join } from "path";
+import {
+  formatTurnsFromData,
+  groupTurnsNaturally,
+  formatGroupedContent,
+  detectContentType,
+  formatResultContent,
+  formatToolWithResult,
+  type Turn,
+  type ToolUse,
+  type ToolResult,
+} from "../src/entrypoints/format-turns";
+
+describe("detectContentType", () => {
+  test("detects JSON objects", () => {
+    expect(detectContentType('{"key": "value"}')).toBe("json");
+    expect(detectContentType('{"number": 42}')).toBe("json");
+  });
+
+  test("detects JSON arrays", () => {
+    expect(detectContentType("[1, 2, 3]")).toBe("json");
+    expect(detectContentType('["a", "b"]')).toBe("json");
+  });
+
+  test("detects Python code", () => {
+    expect(detectContentType("def hello():\n    pass")).toBe("python");
+    expect(detectContentType("import os")).toBe("python");
+    expect(detectContentType("from math import pi")).toBe("python");
+  });
+
+  test("detects JavaScript code", () => {
+    expect(detectContentType("function test() {}")).toBe("javascript");
+    expect(detectContentType("const x = 5")).toBe("javascript");
+    expect(detectContentType("let y = 10")).toBe("javascript");
+    expect(detectContentType("const fn = () => console.log()")).toBe(
+      "javascript",
+    );
+  });
+
+  test("detects bash/shell content", () => {
+    expect(detectContentType("/usr/bin/test")).toBe("bash");
+    expect(detectContentType("Error: command not found")).toBe("bash");
+    expect(detectContentType("ls -la")).toBe("bash");
+    expect(detectContentType("$ echo hello")).toBe("bash");
+  });
+
+  test("detects diff format", () => {
+    expect(detectContentType("@@ -1,3 +1,3 @@")).toBe("diff");
+    expect(detectContentType("+++ file.txt")).toBe("diff");
+    expect(detectContentType("--- file.txt")).toBe("diff");
+  });
+
+  test("detects HTML/XML", () => {
+    expect(detectContentType("<div>hello</div>")).toBe("html");
+    expect(detectContentType("<xml>content</xml>")).toBe("html");
+  });
+
+  test("detects markdown", () => {
+    expect(detectContentType("- List item")).toBe("markdown");
+    expect(detectContentType("* List item")).toBe("markdown");
+    expect(detectContentType("```code```")).toBe("markdown");
+  });
+
+  test("defaults to text", () => {
+    expect(detectContentType("plain text")).toBe("text");
+    expect(detectContentType("just some words")).toBe("text");
+  });
+});
+
+describe("formatResultContent", () => {
+  test("handles empty content", () => {
+    expect(formatResultContent("")).toBe("*(No output)*\n\n");
+    expect(formatResultContent(null)).toBe("*(No output)*\n\n");
+    expect(formatResultContent(undefined)).toBe("*(No output)*\n\n");
+  });
+
+  test("formats short text without code blocks", () => {
+    const result = formatResultContent("success");
+    expect(result).toBe("**→** success\n\n");
+  });
+
+  test("formats long text with code blocks", () => {
+    const longText =
+      "This is a longer piece of text that should be formatted in a code block because it exceeds the short text threshold";
+    const result = formatResultContent(longText);
+    expect(result).toContain("**Result:**");
+    expect(result).toContain("```text");
+    expect(result).toContain(longText);
+  });
+
+  test("pretty prints JSON content", () => {
+    const jsonContent = '{"key": "value", "number": 42}';
+    const result = formatResultContent(jsonContent);
+    expect(result).toContain("```json");
+    expect(result).toContain('"key": "value"');
+    expect(result).toContain('"number": 42');
+  });
+
+  test("truncates very long content", () => {
+    const veryLongContent = "A".repeat(4000);
+    const result = formatResultContent(veryLongContent);
+    expect(result).toContain("...");
+    // Should not contain the full long content
+    expect(result.length).toBeLessThan(veryLongContent.length);
+  });
+
+  test("handles type:text structure", () => {
+    const structuredContent = [{ type: "text", text: "Hello world" }];
+    const result = formatResultContent(JSON.stringify(structuredContent));
+    expect(result).toBe("**→** Hello world\n\n");
+  });
+});
+
+describe("formatToolWithResult", () => {
+  test("formats tool with parameters and result", () => {
+    const toolUse: ToolUse = {
+      type: "tool_use",
+      name: "read_file",
+      input: { file_path: "/path/to/file.txt" },
+      id: "tool_123",
+    };
+
+    const toolResult: ToolResult = {
+      type: "tool_result",
+      tool_use_id: "tool_123",
+      content: "File content here",
+      is_error: false,
+    };
+
+    const result = formatToolWithResult(toolUse, toolResult);
+
+    expect(result).toContain("### 🔧 `read_file`");
+    expect(result).toContain("**Parameters:**");
+    expect(result).toContain('"file_path": "/path/to/file.txt"');
+    expect(result).toContain("**→** File content here");
+  });
+
+  test("formats tool with error result", () => {
+    const toolUse: ToolUse = {
+      type: "tool_use",
+      name: "failing_tool",
+      input: { param: "value" },
+    };
+
+    const toolResult: ToolResult = {
+      type: "tool_result",
+      content: "Permission denied",
+      is_error: true,
+    };
+
+    const result = formatToolWithResult(toolUse, toolResult);
+
+    expect(result).toContain("### 🔧 `failing_tool`");
+    expect(result).toContain("❌ **Error:** `Permission denied`");
+  });
+
+  test("formats tool without parameters", () => {
+    const toolUse: ToolUse = {
+      type: "tool_use",
+      name: "simple_tool",
+    };
+
+    const result = formatToolWithResult(toolUse);
+
+    expect(result).toContain("### 🔧 `simple_tool`");
+    expect(result).not.toContain("**Parameters:**");
+  });
+
+  test("handles unknown tool name", () => {
+    const toolUse: ToolUse = {
+      type: "tool_use",
+    };
+
+    const result = formatToolWithResult(toolUse);
+
+    expect(result).toContain("### 🔧 `unknown_tool`");
+  });
+});
+
+describe("groupTurnsNaturally", () => {
+  test("groups system initialization", () => {
+    const data: Turn[] = [
+      {
+        type: "system",
+        subtype: "init",
+        tools: [{ name: "tool1" }, { name: "tool2" }],
+      },
+    ];
+
+    const result = groupTurnsNaturally(data);
+
+    expect(result).toHaveLength(1);
+    expect(result[0]?.type).toBe("system_init");
+    expect(result[0]?.tools_count).toBe(2);
+  });
+
+  test("groups assistant actions with tool calls", () => {
+    const data: Turn[] = [
+      {
+        type: "assistant",
+        message: {
+          content: [
+            { type: "text", text: "I'll help you" },
+            {
+              type: "tool_use",
+              id: "tool_123",
+              name: "read_file",
+              input: { file_path: "/test.txt" },
+            },
+          ],
+          usage: { input_tokens: 100, output_tokens: 50 },
+        },
+      },
+      {
+        type: "user",
+        message: {
+          content: [
+            {
+              type: "tool_result",
+              tool_use_id: "tool_123",
+              content: "file content",
+              is_error: false,
+            },
+          ],
+        },
+      },
+    ];
+
+    const result = groupTurnsNaturally(data);
+
+    expect(result).toHaveLength(1);
+    expect(result[0]?.type).toBe("assistant_action");
+    expect(result[0]?.text_parts).toEqual(["I'll help you"]);
+    expect(result[0]?.tool_calls).toHaveLength(1);
+    expect(result[0]?.tool_calls?.[0]?.tool_use.name).toBe("read_file");
+    expect(result[0]?.tool_calls?.[0]?.tool_result?.content).toBe(
+      "file content",
+    );
+    expect(result[0]?.usage).toEqual({ input_tokens: 100, output_tokens: 50 });
+  });
+
+  test("groups user messages", () => {
+    const data: Turn[] = [
+      {
+        type: "user",
+        message: {
+          content: [{ type: "text", text: "Please help me" }],
+        },
+      },
+    ];
+
+    const result = groupTurnsNaturally(data);
+
+    expect(result).toHaveLength(1);
+    expect(result[0]?.type).toBe("user_message");
+    expect(result[0]?.text_parts).toEqual(["Please help me"]);
+  });
+
+  test("groups final results", () => {
+    const data: Turn[] = [
+      {
+        type: "result",
+        cost_usd: 0.1234,
+        duration_ms: 5000,
+        result: "Task completed",
+      },
+    ];
+
+    const result = groupTurnsNaturally(data);
+
+    expect(result).toHaveLength(1);
+    expect(result[0]?.type).toBe("final_result");
+    expect(result[0]?.data).toEqual(data[0]!);
+  });
+});
+
+describe("formatGroupedContent", () => {
+  test("formats system initialization", () => {
+    const groupedContent = [
+      {
+        type: "system_init",
+        tools_count: 3,
+      },
+    ];
+
+    const result = formatGroupedContent(groupedContent);
+
+    expect(result).toContain("## Claude Code Report");
+    expect(result).toContain("## 🚀 System Initialization");
+    expect(result).toContain("**Available Tools:** 3 tools loaded");
+  });
+
+  test("formats assistant actions", () => {
+    const groupedContent = [
+      {
+        type: "assistant_action",
+        text_parts: ["I'll help you with that"],
+        tool_calls: [
+          {
+            tool_use: {
+              type: "tool_use",
+              name: "test_tool",
+              input: { param: "value" },
+            },
+            tool_result: {
+              type: "tool_result",
+              content: "result",
+              is_error: false,
+            },
+          },
+        ],
+        usage: { input_tokens: 100, output_tokens: 50 },
+      },
+    ];
+
+    const result = formatGroupedContent(groupedContent);
+
+    expect(result).toContain("I'll help you with that");
+    expect(result).toContain("### 🔧 `test_tool`");
+    expect(result).toContain("*Token usage: 100 input, 50 output*");
+  });
+
+  test("formats user messages", () => {
+    const groupedContent = [
+      {
+        type: "user_message",
+        text_parts: ["Help me please"],
+      },
+    ];
+
+    const result = formatGroupedContent(groupedContent);
+
+    expect(result).toContain("## 👤 User");
+    expect(result).toContain("Help me please");
+  });
+
+  test("formats final results", () => {
+    const groupedContent = [
+      {
+        type: "final_result",
+        data: {
+          type: "result",
+          cost_usd: 0.1234,
+          duration_ms: 5678,
+          result: "Success!",
+        } as Turn,
+      },
+    ];
+
+    const result = formatGroupedContent(groupedContent);
+
+    expect(result).toContain("## ✅ Final Result");
+    expect(result).toContain("Success!");
+    expect(result).toContain("**Cost:** $0.1234");
+    expect(result).toContain("**Duration:** 5.7s");
+  });
+});
+
+describe("formatTurnsFromData", () => {
+  test("handles empty data", () => {
+    const result = formatTurnsFromData([]);
+    expect(result).toBe("## Claude Code Report\n\n");
+  });
+
+  test("formats complete conversation", () => {
+    const data: Turn[] = [
+      {
+        type: "system",
+        subtype: "init",
+        tools: [{ name: "tool1" }],
+      },
+      {
+        type: "assistant",
+        message: {
+          content: [
+            { type: "text", text: "I'll help you" },
+            {
+              type: "tool_use",
+              id: "tool_123",
+              name: "read_file",
+              input: { file_path: "/test.txt" },
+            },
+          ],
+        },
+      },
+      {
+        type: "user",
+        message: {
+          content: [
+            {
+              type: "tool_result",
+              tool_use_id: "tool_123",
+              content: "file content",
+              is_error: false,
+            },
+          ],
+        },
+      },
+      {
+        type: "result",
+        cost_usd: 0.05,
+        duration_ms: 2000,
+        result: "Done",
+      },
+    ];
+
+    const result = formatTurnsFromData(data);
+
+    expect(result).toContain("## Claude Code Report");
+    expect(result).toContain("## 🚀 System Initialization");
+    expect(result).toContain("I'll help you");
+    expect(result).toContain("### 🔧 `read_file`");
+    expect(result).toContain("## ✅ Final Result");
+    expect(result).toContain("Done");
+  });
+});
+
+describe("integration tests", () => {
+  test("formats real conversation data correctly", () => {
+    // Load the sample JSON data
+    const jsonPath = join(__dirname, "fixtures", "sample-turns.json");
+    const expectedPath = join(
+      __dirname,
+      "fixtures",
+      "sample-turns-expected-output.md",
+    );
+
+    const jsonData = JSON.parse(readFileSync(jsonPath, "utf-8"));
+    const expectedOutput = readFileSync(expectedPath, "utf-8").trim();
+
+    // Format the data using our function
+    const actualOutput = formatTurnsFromData(jsonData).trim();
+
+    // Compare the outputs
+    expect(actualOutput).toBe(expectedOutput);
+  });
+});

test/github/context.test.ts

@@ -1,5 +1,8 @@
 import { describe, it, expect } from "bun:test";
-import { parseMultilineInput } from "../../src/github/context";
+import {
+  parseMultilineInput,
+  parseAdditionalPermissions,
+} from "../../src/github/context";
 
 describe("parseMultilineInput", () => {
   it("should parse a comma-separated string", () => {
@@ -55,3 +58,58 @@ Bash(bun typecheck)
     expect(result).toEqual([]);
   });
 });
+
+describe("parseAdditionalPermissions", () => {
+  it("should parse single permission", () => {
+    const input = "actions: read";
+    const result = parseAdditionalPermissions(input);
+    expect(result.get("actions")).toBe("read");
+    expect(result.size).toBe(1);
+  });
+
+  it("should parse multiple permissions", () => {
+    const input = `actions: read
+packages: write
+contents: read`;
+    const result = parseAdditionalPermissions(input);
+    expect(result.get("actions")).toBe("read");
+    expect(result.get("packages")).toBe("write");
+    expect(result.get("contents")).toBe("read");
+    expect(result.size).toBe(3);
+  });
+
+  it("should handle empty string", () => {
+    const input = "";
+    const result = parseAdditionalPermissions(input);
+    expect(result.size).toBe(0);
+  });
+
+  it("should handle whitespace and empty lines", () => {
+    const input = `
+    actions: read
+
+    packages: write
+    `;
+    const result = parseAdditionalPermissions(input);
+    expect(result.get("actions")).toBe("read");
+    expect(result.get("packages")).toBe("write");
+    expect(result.size).toBe(2);
+  });
+
+  it("should ignore lines without colon separator", () => {
+    const input = `actions: read
+invalid line
+packages: write`;
+    const result = parseAdditionalPermissions(input);
+    expect(result.get("actions")).toBe("read");
+    expect(result.get("packages")).toBe("write");
+    expect(result.size).toBe(2);
+  });
+
+  it("should trim whitespace around keys and values", () => {
+    const input = "  actions  :  read  ";
+    const result = parseAdditionalPermissions(input);
+    expect(result.get("actions")).toBe("read");
+    expect(result.size).toBe(1);
+  });
+});

test/install-mcp-server.test.ts

@@ -1,6 +1,7 @@
 import { describe, test, expect, beforeEach, afterEach, spyOn } from "bun:test";
 import { prepareMcpConfig } from "../src/mcp/install-mcp-server";
 import * as core from "@actions/core";
+import type { ParsedGitHubContext } from "../src/github/context";
 
 describe("prepareMcpConfig", () => {
   let consoleInfoSpy: any;
@@ -8,6 +9,41 @@ describe("prepareMcpConfig", () => {
   let setFailedSpy: any;
   let processExitSpy: any;
 
+  // Create a mock context for tests
+  const mockContext: ParsedGitHubContext = {
+    runId: "test-run-id",
+    eventName: "issue_comment",
+    eventAction: "created",
+    repository: {
+      owner: "test-owner",
+      repo: "test-repo",
+      full_name: "test-owner/test-repo",
+    },
+    actor: "test-actor",
+    payload: {} as any,
+    entityNumber: 123,
+    isPR: false,
+    inputs: {
+      triggerPhrase: "@claude",
+      assigneeTrigger: "",
+      labelTrigger: "",
+      allowedTools: [],
+      disallowedTools: [],
+      customInstructions: "",
+      directPrompt: "",
+      branchPrefix: "",
+      useStickyComment: false,
+      additionalPermissions: new Map(),
+    },
+  };
+
+  const mockPRContext: ParsedGitHubContext = {
+    ...mockContext,
+    eventName: "pull_request",
+    isPR: true,
+    entityNumber: 456,
+  };
+
   beforeEach(() => {
     consoleInfoSpy = spyOn(core, "info").mockImplementation(() => {});
     consoleWarningSpy = spyOn(core, "warning").mockImplementation(() => {});
@@ -15,6 +51,11 @@ describe("prepareMcpConfig", () => {
     processExitSpy = spyOn(process, "exit").mockImplementation(() => {
       throw new Error("Process exit");
     });
+
+    // Set up required environment variables
+    if (!process.env.GITHUB_ACTION_PATH) {
+      process.env.GITHUB_ACTION_PATH = "/test/action/path";
+    }
   });
 
   afterEach(() => {
@@ -31,6 +72,7 @@ describe("prepareMcpConfig", () => {
       repo: "test-repo",
       branch: "test-branch",
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -57,6 +99,7 @@ describe("prepareMcpConfig", () => {
         "mcp__github__create_issue",
         "mcp__github_file_ops__commit_files",
       ],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -78,6 +121,7 @@ describe("prepareMcpConfig", () => {
         "mcp__github_file_ops__commit_files",
         "mcp__github_file_ops__update_claude_comment",
       ],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -93,6 +137,7 @@ describe("prepareMcpConfig", () => {
       repo: "test-repo",
       branch: "test-branch",
       allowedTools: ["Edit", "Read", "Write"],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -109,6 +154,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: "",
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -126,6 +172,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: "   \n\t  ",
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -158,6 +205,7 @@ describe("prepareMcpConfig", () => {
         "mcp__github__create_issue",
         "mcp__github_file_ops__commit_files",
       ],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -195,6 +243,7 @@ describe("prepareMcpConfig", () => {
         "mcp__github__create_issue",
         "mcp__github_file_ops__commit_files",
       ],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -232,6 +281,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: additionalConfig,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -251,6 +301,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: invalidJson,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -271,6 +322,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: nonObjectJson,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -294,6 +346,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: nullJson,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -317,6 +370,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: arrayJson,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -363,6 +417,7 @@ describe("prepareMcpConfig", () => {
       branch: "test-branch",
       additionalMcpConfig: additionalConfig,
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -384,6 +439,7 @@ describe("prepareMcpConfig", () => {
       repo: "test-repo",
       branch: "test-branch",
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -404,6 +460,7 @@ describe("prepareMcpConfig", () => {
       repo: "test-repo",
       branch: "test-branch",
       allowedTools: [],
+      context: mockContext,
     });
 
     const parsed = JSON.parse(result);
@@ -411,4 +468,132 @@ describe("prepareMcpConfig", () => {
 
     process.env.GITHUB_WORKSPACE = oldEnv;
   });
+
+  test("should include github_ci server when context.isPR is true and actions:read permission is granted", async () => {
+    const oldEnv = process.env.ACTIONS_TOKEN;
+    process.env.ACTIONS_TOKEN = "workflow-token";
+
+    const contextWithPermissions = {
+      ...mockPRContext,
+      inputs: {
+        ...mockPRContext.inputs,
+        additionalPermissions: new Map([["actions", "read"]]),
+      },
+    };
+
+    const result = await prepareMcpConfig({
+      githubToken: "test-token",
+      owner: "test-owner",
+      repo: "test-repo",
+      branch: "test-branch",
+      allowedTools: [],
+      context: contextWithPermissions,
+    });
+
+    const parsed = JSON.parse(result);
+    expect(parsed.mcpServers.github_ci).toBeDefined();
+    expect(parsed.mcpServers.github_ci.env.GITHUB_TOKEN).toBe("workflow-token");
+    expect(parsed.mcpServers.github_ci.env.PR_NUMBER).toBe("456");
+    expect(parsed.mcpServers.github_file_ops).toBeDefined();
+
+    process.env.ACTIONS_TOKEN = oldEnv;
+  });
+
+  test("should not include github_ci server when context.isPR is false", async () => {
+    const result = await prepareMcpConfig({
+      githubToken: "test-token",
+      owner: "test-owner",
+      repo: "test-repo",
+      branch: "test-branch",
+      allowedTools: [],
+      context: mockContext,
+    });
+
+    const parsed = JSON.parse(result);
+    expect(parsed.mcpServers.github_ci).not.toBeDefined();
+    expect(parsed.mcpServers.github_file_ops).toBeDefined();
+  });
+
+  test("should not include github_ci server when actions:read permission is not granted", async () => {
+    const oldTokenEnv = process.env.ACTIONS_TOKEN;
+    process.env.ACTIONS_TOKEN = "workflow-token";
+
+    const result = await prepareMcpConfig({
+      githubToken: "test-token",
+      owner: "test-owner",
+      repo: "test-repo",
+      branch: "test-branch",
+      allowedTools: [],
+      context: mockPRContext,
+    });
+
+    const parsed = JSON.parse(result);
+    expect(parsed.mcpServers.github_ci).not.toBeDefined();
+    expect(parsed.mcpServers.github_file_ops).toBeDefined();
+
+    process.env.ACTIONS_TOKEN = oldTokenEnv;
+  });
+
+  test("should parse additional_permissions with multiple lines correctly", async () => {
+    const oldTokenEnv = process.env.ACTIONS_TOKEN;
+    process.env.ACTIONS_TOKEN = "workflow-token";
+
+    const contextWithPermissions = {
+      ...mockPRContext,
+      inputs: {
+        ...mockPRContext.inputs,
+        additionalPermissions: new Map([
+          ["actions", "read"],
+          ["future", "permission"],
+        ]),
+      },
+    };
+
+    const result = await prepareMcpConfig({
+      githubToken: "test-token",
+      owner: "test-owner",
+      repo: "test-repo",
+      branch: "test-branch",
+      allowedTools: [],
+      context: contextWithPermissions,
+    });
+
+    const parsed = JSON.parse(result);
+    expect(parsed.mcpServers.github_ci).toBeDefined();
+    expect(parsed.mcpServers.github_ci.env.GITHUB_TOKEN).toBe("workflow-token");
+
+    process.env.ACTIONS_TOKEN = oldTokenEnv;
+  });
+
+  test("should warn when actions:read is requested but token lacks permission", async () => {
+    const oldTokenEnv = process.env.ACTIONS_TOKEN;
+    process.env.ACTIONS_TOKEN = "invalid-token";
+
+    const contextWithPermissions = {
+      ...mockPRContext,
+      inputs: {
+        ...mockPRContext.inputs,
+        additionalPermissions: new Map([["actions", "read"]]),
+      },
+    };
+
+    const result = await prepareMcpConfig({
+      githubToken: "test-token",
+      owner: "test-owner",
+      repo: "test-repo",
+      branch: "test-branch",
+      allowedTools: [],
+      context: contextWithPermissions,
+    });
+
+    const parsed = JSON.parse(result);
+    expect(parsed.mcpServers.github_ci).toBeDefined();
+    expect(consoleWarningSpy).toHaveBeenCalledWith(
+      expect.stringContaining(
+        "The github_ci MCP server requires 'actions: read' permission",
+      ),
+    );
+
+    process.env.ACTIONS_TOKEN = oldTokenEnv;
+  });
 });

test/mockContext.ts

@@ -10,6 +10,7 @@ import type {
 const defaultInputs = {
   triggerPhrase: "/claude",
   assigneeTrigger: "",
+  labelTrigger: "",
   anthropicModel: "claude-3-7-sonnet-20250219",
   allowedTools: [] as string[],
   disallowedTools: [] as string[],
@@ -18,6 +19,9 @@ const defaultInputs = {
   useBedrock: false,
   useVertex: false,
   timeoutMinutes: 30,
+  branchPrefix: "claude/",
+  useStickyComment: false,
+  additionalPermissions: new Map<string, string>(),
 };
 
 const defaultRepository = {
@@ -128,6 +132,46 @@ export const mockIssueAssignedContext: ParsedGitHubContext = {
   inputs: { ...defaultInputs, assigneeTrigger: "@claude-bot" },
 };
 
+export const mockIssueLabeledContext: ParsedGitHubContext = {
+  runId: "1234567890",
+  eventName: "issues",
+  eventAction: "labeled",
+  repository: defaultRepository,
+  actor: "admin-user",
+  payload: {
+    action: "labeled",
+    issue: {
+      number: 1234,
+      title: "Enhancement: Improve search functionality",
+      body: "The current search is too slow and needs optimization",
+      user: {
+        login: "alice-wonder",
+        id: 54321,
+        avatar_url: "https://avatars.githubusercontent.com/u/54321",
+        html_url: "https://github.com/alice-wonder",
+      },
+      assignee: null,
+    },
+    label: {
+      id: 987654321,
+      name: "claude-task",
+      color: "f29513",
+      description: "Label for Claude AI interactions",
+    },
+    repository: {
+      name: "test-repo",
+      full_name: "test-owner/test-repo",
+      private: false,
+      owner: {
+        login: "test-owner",
+      },
+    },
+  } as IssuesEvent,
+  entityNumber: 1234,
+  isPR: false,
+  inputs: { ...defaultInputs, labelTrigger: "claude-task" },
+};
+
 // Issue comment on issue event
 export const mockIssueCommentContext: ParsedGitHubContext = {
   runId: "1234567890",

test/permissions.test.ts

@@ -62,10 +62,14 @@ describe("checkWritePermissions", () => {
     inputs: {
       triggerPhrase: "@claude",
       assigneeTrigger: "",
+      labelTrigger: "",
       allowedTools: [],
       disallowedTools: [],
       customInstructions: "",
       directPrompt: "",
+      branchPrefix: "claude/",
+      useStickyComment: false,
+      additionalPermissions: new Map(),
     },
   });
 

test/prepare-context.test.ts

@@ -219,6 +219,55 @@ describe("parseEnvVarsWithContext", () => {
         ),
       ).toThrow("BASE_BRANCH is required for issues event");
     });
+
+    test("should allow issue assigned event with direct_prompt and no assigneeTrigger", () => {
+      const contextWithDirectPrompt = createMockContext({
+        ...mockIssueAssignedContext,
+        inputs: {
+          ...mockIssueAssignedContext.inputs,
+          assigneeTrigger: "", // No assignee trigger
+          directPrompt: "Please assess this issue", // But direct prompt is provided
+        },
+      });
+
+      const result = prepareContext(
+        contextWithDirectPrompt,
+        "12345",
+        "main",
+        "claude/issue-123-20240101_120000",
+      );
+
+      expect(result.eventData.eventName).toBe("issues");
+      expect(result.eventData.isPR).toBe(false);
+      expect(result.directPrompt).toBe("Please assess this issue");
+      if (
+        result.eventData.eventName === "issues" &&
+        result.eventData.eventAction === "assigned"
+      ) {
+        expect(result.eventData.issueNumber).toBe("123");
+        expect(result.eventData.assigneeTrigger).toBeUndefined();
+      }
+    });
+
+    test("should throw error when neither assigneeTrigger nor directPrompt provided for issue assigned event", () => {
+      const contextWithoutTriggers = createMockContext({
+        ...mockIssueAssignedContext,
+        inputs: {
+          ...mockIssueAssignedContext.inputs,
+          assigneeTrigger: "", // No assignee trigger
+          directPrompt: "", // No direct prompt
+        },
+      });
+
+      expect(() =>
+        prepareContext(
+          contextWithoutTriggers,
+          "12345",
+          "main",
+          "claude/issue-123-20240101_120000",
+        ),
+      ).toThrow("ASSIGNEE_TRIGGER is required for issue assigned event");
+    });
   });
 
   describe("optional fields", () => {

test/trigger-validation.test.ts

@@ -6,6 +6,7 @@ import { describe, it, expect } from "bun:test";
 import {
   createMockContext,
   mockIssueAssignedContext,
+  mockIssueLabeledContext,
   mockIssueCommentContext,
   mockIssueOpenedContext,
   mockPullRequestReviewContext,
@@ -29,10 +30,14 @@ describe("checkContainsTrigger", () => {
         inputs: {
           triggerPhrase: "/claude",
           assigneeTrigger: "",
+          labelTrigger: "",
           directPrompt: "Fix the bug in the login form",
           allowedTools: [],
           disallowedTools: [],
           customInstructions: "",
+          branchPrefix: "claude/",
+          useStickyComment: false,
+          additionalPermissions: new Map(),
         },
       });
       expect(checkContainsTrigger(context)).toBe(true);
@@ -55,10 +60,14 @@ describe("checkContainsTrigger", () => {
         inputs: {
           triggerPhrase: "/claude",
           assigneeTrigger: "",
+          labelTrigger: "",
           directPrompt: "",
           allowedTools: [],
           disallowedTools: [],
           customInstructions: "",
+          branchPrefix: "claude/",
+          useStickyComment: false,
+          additionalPermissions: new Map(),
         },
       });
       expect(checkContainsTrigger(context)).toBe(false);
@@ -107,6 +116,39 @@ describe("checkContainsTrigger", () => {
     });
   });
 
+  describe("label trigger", () => {
+    it("should return true when issue is labeled with the trigger label", () => {
+      const context = mockIssueLabeledContext;
+      expect(checkContainsTrigger(context)).toBe(true);
+    });
+
+    it("should return false when issue is labeled with a different label", () => {
+      const context = {
+        ...mockIssueLabeledContext,
+        payload: {
+          ...mockIssueLabeledContext.payload,
+          label: {
+            ...(mockIssueLabeledContext.payload as any).label,
+            name: "bug",
+          },
+        },
+      } as ParsedGitHubContext;
+      expect(checkContainsTrigger(context)).toBe(false);
+    });
+
+    it("should return false for non-labeled events", () => {
+      const context = {
+        ...mockIssueLabeledContext,
+        eventAction: "opened",
+        payload: {
+          ...mockIssueLabeledContext.payload,
+          action: "opened",
+        },
+      } as ParsedGitHubContext;
+      expect(checkContainsTrigger(context)).toBe(false);
+    });
+  });
+
   describe("issue body and title trigger", () => {
     it("should return true when issue body contains trigger phrase", () => {
       const context = mockIssueOpenedContext;
@@ -232,10 +274,14 @@ describe("checkContainsTrigger", () => {
         inputs: {
           triggerPhrase: "@claude",
           assigneeTrigger: "",
+          labelTrigger: "",
           directPrompt: "",
           allowedTools: [],
           disallowedTools: [],
           customInstructions: "",
+          branchPrefix: "claude/",
+          useStickyComment: false,
+          additionalPermissions: new Map(),
         },
       });
       expect(checkContainsTrigger(context)).toBe(true);
@@ -259,10 +305,14 @@ describe("checkContainsTrigger", () => {
         inputs: {
           triggerPhrase: "@claude",
           assigneeTrigger: "",
+          labelTrigger: "",
           directPrompt: "",
           allowedTools: [],
           disallowedTools: [],
           customInstructions: "",
+          branchPrefix: "claude/",
+          useStickyComment: false,
+          additionalPermissions: new Map(),
         },
       });
       expect(checkContainsTrigger(context)).toBe(true);
@@ -286,10 +336,14 @@ describe("checkContainsTrigger", () => {
         inputs: {
           triggerPhrase: "@claude",
           assigneeTrigger: "",
+          labelTrigger: "",
           directPrompt: "",
           allowedTools: [],
           disallowedTools: [],
           customInstructions: "",
+          branchPrefix: "claude/",
+          useStickyComment: false,
+          additionalPermissions: new Map(),
         },
       });
       expect(checkContainsTrigger(context)).toBe(false);