Skip to content

Let ACP validation choose PR test plans#189

Open
osolmaz wants to merge 2 commits intomainfrom
codex/pr-triage-model-driven-validation
Open

Let ACP validation choose PR test plans#189
osolmaz wants to merge 2 commits intomainfrom
codex/pr-triage-model-driven-validation

Conversation

@osolmaz
Copy link
Copy Markdown
Contributor

@osolmaz osolmaz commented Mar 28, 2026

Summary

  • move the bug and feature validation nodes back into ACP so the model chooses and runs the validation plan itself
  • remove the hardcoded runtime helpers that inferred tests from changed test files and auto-ran installs
  • update the PR-triage README and regression test so the executable flow and markdown stay aligned

Validation

  • pnpm run build:test && node --test dist-test/test/pr-triage-example.test.js
  • pnpm run check:docs
  • pnpm run check

Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3a59842f95

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +797 to +799
' "route": "judge_refactor" | "comment_and_escalate_to_human",',
' "validation_result": "validated" | "blocked" | "not_proven",',
' "summary": "short explanation",',
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Prevent blocked validation from routing to judge_refactor

This schema now allows contradictory outputs where validation is not proven but the flow still advances, e.g. {"route":"judge_refactor","validation_result":"blocked"}. Because the workflow switches only on route, a malformed ACP response can bypass the validation gate and continue into refactor/review with an unproven fix. The previous action-based implementation always forced non-proven outcomes to escalation, so this is a regression in safety for autonomous triage (the same pattern is also present in the feature-validation prompt).

Useful? React with 👍 / 👎.

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.

1 participant