Lighten WATCHLIST skill and validation - #32
Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b955dbee58
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".
| "PASSWORD_ASSIGNMENT": ( | ||
| r"\b(?:password|passwd|pwd|api[_-]?key|secret[_-]?key)\s*[:=]\s*\S+" | ||
| ), |
There was a problem hiding this comment.
Reject access-token assignments
When a watchlist contains a credential written as access_token: <value> or access-token=<value>, the new assignment pattern does not match it, so the validator reports success even though the security contract forbids storing tokens. The previous API_KEY_ASSIGNMENT pattern explicitly covered access[_-]?token; retain that alternative in the consolidated pattern to avoid allowing these secrets into validated files.
Useful? React with 👍 / 👎.
Summary
.watchlist/WATCHLIST.mdSKILL.mdand simplify validator/tests/docs##section boundaryWhy
Compatibility layers, repeated reference documents, archive/automation policy, and duplicated semantic corpora made the package larger and introduced conflicting storage, routing, and time semantics. This change keeps the effective file workflow while applying YAGNI to unused policy and keeping privacy, safety, and deterministic validation explicit.
Validation
python -B -m unittest discover -s evals -p 'test_*.py'— 45 passedgit diff --checkpassedfeb95262949498689ab4d3e4baec255489858ed061bb118abbe187f08e917584add-private-default,existing-timezone-authoritative,review-read-only,complete-user-reported,cross-target-duplicate-stop,unsupported-schema-stops-before-side-effects,negative-generic-reminder,negative-generic-lifecycle$watchlist-mdinvocationThe manual runtime corpus now contains 20 cases. The remaining 12 cases are pending and are not presented as complete.