claude-code-action/src/modes/agent/index.ts

import * as core from "@actions/core";
import { mkdir, writeFile } from "fs/promises";
import type { Mode, ModeOptions, ModeResult } from "../types";
import { isAutomationContext } from "../../github/context";

/**
 * Agent mode implementation.
 *
 * This mode is specifically designed for automation events (workflow_dispatch and schedule).
 * It bypasses the standard trigger checking and comment tracking used by tag mode,
 * making it ideal for scheduled tasks and manual workflow runs.
 */
export const agentMode: Mode = {
  name: "agent",
  description: "Automation mode for workflow_dispatch and schedule events",

  shouldTrigger(context) {
    // Only trigger for automation events
    return isAutomationContext(context);
  },

  prepareContext(context) {
    // Agent mode doesn't use comment tracking or branch management
    return {
      mode: "agent",
      githubContext: context,
    };
  },

  getAllowedTools() {
    return [];
  },

  getDisallowedTools() {
    return [];
  },

  shouldCreateTrackingComment() {
    return false;
  },

  async prepare({ context }: ModeOptions): Promise<ModeResult> {
    // Agent mode handles automation events (workflow_dispatch, schedule) only

    // Create prompt directory
    await mkdir(`${process.env.RUNNER_TEMP}/claude-prompts`, {
      recursive: true,
    });

    // Write the prompt file - the base action requires a prompt_file parameter,
    // so we must create this file even though agent mode typically uses
    // override_prompt or direct_prompt. If neither is provided, we write
    // a minimal prompt with just the repository information.
    const promptContent =
      context.inputs.overridePrompt ||
      context.inputs.directPrompt ||
      `Repository: ${context.repository.owner}/${context.repository.repo}`;

    await writeFile(
      `${process.env.RUNNER_TEMP}/claude-prompts/claude-prompt.txt`,
      promptContent,
    );

    // Export tool environment variables for agent mode
    const baseTools = [
      "Edit",
      "MultiEdit",
      "Glob",
      "Grep",
      "LS",
      "Read",
      "Write",
    ];

    // Add user-specified tools
    const allowedTools = [...baseTools, ...context.inputs.allowedTools];
    const disallowedTools = [
      "WebSearch",
      "WebFetch",
      ...context.inputs.disallowedTools,
    ];

    core.exportVariable("ALLOWED_TOOLS", allowedTools.join(","));
    core.exportVariable("DISALLOWED_TOOLS", disallowedTools.join(","));

    // Agent mode uses a minimal MCP configuration
    // We don't need comment servers or PR-specific tools for automation
    const mcpConfig: any = {
      mcpServers: {},
    };

    // Add user-provided additional MCP config if any
    const additionalMcpConfig = process.env.MCP_CONFIG || "";
    if (additionalMcpConfig.trim()) {
      try {
        const additional = JSON.parse(additionalMcpConfig);
        if (additional && typeof additional === "object") {
          Object.assign(mcpConfig, additional);
        }
      } catch (error) {
        core.warning(`Failed to parse additional MCP config: ${error}`);
      }
    }

    core.setOutput("mcp_config", JSON.stringify(mcpConfig));

    return {
      commentId: undefined,
      branchInfo: {
        baseBranch: "",
        currentBranch: "",
        claudeBranch: undefined,
      },
      mcpConfig: JSON.stringify(mcpConfig),
    };
  },
};