--- description: Collaborate with another agent via a shared markdown file argument-hint: [initial task description] --- # Collaboration Session You are collaborating with another agent. Coordinate via a shared markdown file. ## Task $ARGUMENTS ## Setup 1. Create a shared collaboration file at `/tmp/collab-.md` with this structure: ```markdown # Collaboration: **Started:** **This agent:** **Other agent:** ## Task ## Context ## Requests for the other agent ## Answers ``` 2. Report the exact file path to the user and give them this message to pass to the other agent: > You are collaborating with another Claude agent via the shared file at ``. > - Read the file to see the current task and requests. > - Append your responses to the `## Answers` section using the `### Answer N (timestamp)` format shown in the file. > - Do not edit sections above `## Answers` — only append. > - When you have new questions or need clarification, add them under a `## Questions for the other agent` section at the bottom. > - If you need the user to clarify something, pause and ask them directly before continuing. ## Watch loop After setup, poll the file for changes: - Check the file's mtime every 5 seconds (use `stat -c %Y ` on Linux). - When the file changes, read the new content and report back to the user with a summary of what the other agent added. - Do not summarize silently — always surface the new content so the user sees the exchange. - Continue watching until the user says the collaboration is complete. ## Clarifications If at any point the task, a response, or a request is ambiguous, stop and ask the user directly before writing to the file or continuing the watch loop. Do not guess on behalf of either agent. ## Rules - Only append to the `## Answers` section from this side when responding to the other agent's questions. - Keep entries timestamped and numbered sequentially. - Never delete prior content — the file is the audit log of the collaboration.