fix: stop guard treats stale task.json as expired after 60 minutes

[nikosbosse]

Summary

• Fixes a bug where a stale ~/.everyrow/task.json with "status": "running" blocks all Claude Code sessions from stopping indefinitely
• The stop guard now checks file mtime (updated on every everyrow_progress poll) — if no poll has happened in 60 minutes, the session is likely dead, so it removes the file and allows the stop
• Using mtime rather than started_at means long-running tasks that are still being actively polled won't be affected
• Root cause: if a session crashes or is force-quit before everyrow_progress polling completes, SessionEnd doesn't always fire, leaving a permanent stale state file

Test plan

• Existing tests updated: test 1 now includes started_at (matches real MCP server output)
• New test 5: stale file mtime (2 hours old) → allows stop and removes file
• New test 6: recent file mtime (just written) → still blocks as expected
• All 6 tests pass: bash everyrow-mcp/tests/test_hook_stop_guard.sh
• Results hook tests still pass: bash everyrow-mcp/tests/test_hook_results.sh

🤖 Generated with Claude Code

[rgambee]

Thanks. I would like to get rid of this stop guard soon, but this is a helpful incremental improvement.

[rgambee]

This could probably be lowered significantly. The file is supposed to update every 15 seconds or so. 5 minutes or even 1 minute of inactivity is likely a sign that it can be ignored.

[nikosbosse]

Changed to 4 minutes. Will merge soon

[sentry]

Bug: The STALE_SECONDS variable is set to 240 (4 minutes), but the documented intent is 60 minutes, which could cause premature termination of active sessions.
_{Severity: CRITICAL}

Suggested Fix

In everyrow-stop-guard.sh, change the value of STALE_SECONDS from 240 to 3600 to align with the documented 60-minute timeout.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: everyrow-mcp/scripts/everyrow-stop-guard.sh#L16

Potential issue: The script `everyrow-stop-guard.sh` sets a timeout of 4 minutes
(`STALE_SECONDS=240`) to determine if a session is stale. However, the PR description
and commit message state the intended timeout is 60 minutes. With polling happening
approximately every 12 seconds, any task experiencing a delay longer than 4 minutes due
to network issues or system load will be incorrectly terminated. This aggressive timeout
contradicts the intended safety margin and will lead to the premature termination of
potentially active user sessions.

_{Did we get this right? 👍 / 👎 to inform future reviews.}