Preserve verification evidence and prerequisite retry budgets - #13
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9270df2250
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".
| unavailable. A verification infrastructure failure preserves the same saved | ||
| candidate and requests evidence retry. A demonstrated code defect requests |
There was a problem hiding this comment.
When an infrastructure failure occurs on the second authorized setup attempt, this unconditionally requests another evidence retry even though the preserved budget is now exhausted. That conflicts with the later Continuation Decision rules, which require an unchanged unavailable prerequisite to terminate with needs_human or blocked rather than repeat the check, so an orchestrator can keep re-entering verification with no permitted recovery action; make evidence retry conditional on remaining budget or a concrete environment change.
Useful? React with 👍 / 👎.
1 participant
Verification continuations can lose the requirement inventory and repeat prerequisite setup without retaining the earlier attempt budget. Require the portable verifier to persist its requirement map against the exact candidate, retain at most two setup attempts per owning service and candidate, and distinguish evidence retry from code remediation and missing authority.
This places the behavior in the owning MoonSpec bundle so every consumer receives the same instructions. It supplies the canonical source for the matching MoonMind reliability change in MoonLadderStudios/MoonMind#4281.
Validation:
python tools/validate_bundle.pypassed.tools/project_bundle.py --projection moonmind --writeand--checkpassed against a disposable Linux target.python tools/sync_moonspec.py --check, passed in Linux with this source and its byte-identical projection.