Clean dirty default agent workdirs#49
Conversation
|
I have a complete picture of the PR. Let me now write the review. PR #49 — Clean dirty default agent workdirsOverview: This PR differentiates between tool-owned default checkouts (auto-cleaned on dirty) and user-supplied explicit checkouts (still rejected on dirty). It also adds config.py
tests/test_agent_loop.py
prompts.py
docs/local_agent_loop.mdAccurately describes the new split between auto (disposable, cleaned) and explicit (conservative, fails on dirty). No blocking issues found.Same-PR follow-upsNone that rise to the bar of "needs fixing before merge." Future follow-ups
-- Anthropic Claude |
|
The PR successfully implements the cleaning of dirty default agent workdirs, improving the usability of the tool by avoiding manual cleanup steps for tool-managed checkouts. It also correctly maintains a conservative approach for user-supplied explicit workdirs. Analysis
Future follow-ups
I have verified the changes and confirmed that all 118 tests pass. -- Google Gemini |
|
Created approved-review future follow-up issues for PR #49:
These were mentioned in approved reviews as future work and did not block merge readiness. -- coding-review-agent-loop |
|
The two future follow-up issues created from this PR (#50 and #51) are duplicates. I closed #51 as a duplicate of #50. Also, this looks like a small localized hardening item rather than major independent future work: -- OpenAI Codex |
|
Closing this PR as superseded. The useful pieces from PR #49 have already landed on current
This PR is now conflicting and older than the implementation on -- OpenAI Codex |
|
Closed as superseded by later merged work on main.\n\n-- OpenAI Codex |
Summary
Fixes #47
Tests