Skip to content

Preserve verification failure evidence - #1890

Merged
chubes4 merged 1 commit into
mainfrom
fix/1889-verification-failure-evidence
Jul 20, 2026
Merged

Preserve verification failure evidence#1890
chubes4 merged 1 commit into
mainfrom
fix/1889-verification-failure-evidence

Conversation

@chubes4

@chubes4 chubes4 commented Jul 20, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • retain bounded sanitized stdout and stderr tails for failed verification and drift commands
  • attach successfully applied patch and changed-file descriptors when later verification fails
  • stage those artifacts through the existing reviewer-safe apply-failure upload policy

Root cause

Docs Agent maintenance runs applied their patches and passed every repository check, then failed the completion validator. WP Codebox projected only the command exit code and discarded both the validator diagnostic and the applied diff, making the failure impossible to classify from GitHub Actions artifacts.

Reproductions:

How to test

  1. Run npm run test:native-agent-task-lifecycle.
  2. Run npm run test:native-agent-task-interruption.
  3. Run npm run test:agent-task-workflow-interface.
  4. Run actionlint .github/workflows/*.yml.
  5. Confirm the lifecycle fixture exposes failed command tails and stages sanitized patch/change evidence without changing successful result projections.

Compatibility

This adds reviewer evidence only on failed verification or drift commands. Successful workflow result projections and artifact paths remain unchanged. No extension interface is removed or renamed.

Closes #1889

AI assistance

  • AI assistance: Yes
  • Tool(s): OpenCode with GPT-5.6-sol
  • Used for: Diagnosed two hosted failures and implemented the generic evidence projection and deterministic coverage with Chris reviewing and owning the change.

@chubes4
chubes4 merged commit 0227ca7 into main Jul 20, 2026
4 checks passed
@chubes4
chubes4 deleted the fix/1889-verification-failure-evidence branch July 20, 2026 11:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Preserve verification failure diagnostics and applied diff evidence

1 participant