Repository navigation
Conversation
The review and adversarial-review commands told Claude not to call `BashOutput`, a tool Claude Code no longer has (background output is now read from the task's output file). Say what the rule means instead: do not read the background output or wait for it this turn. The result-handling skill stacked CRITICAL, STOP, MUST and "strictly forbidden" on the review-only rule. Current Claude models follow plain instructions closely, and the stacked emphasis invites over-application. The rule itself is unchanged: present findings, ask which to fix, make no edits until the user picks.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
commands/review.mdandcommands/adversarial-review.mdsay "Do not callBashOutput". Claude Code no longer has aBashOutputtool (its tools reference has no such entry; background output is read from the task's output file). The line now states the intent directly: do not read the background command's output or wait for completion in this turn.skills/codex-result-handling/SKILL.mdstacks CRITICAL / STOP / MUST / "strictly forbidden" on the review-only rule. Current Claude models follow plain instructions closely, and stacked emphasis tends to over-apply. The rule is unchanged: present findings, ask which to fix, make no edits until the user picks.tests/commands.test.mjs: the two assertions that matched the oldBashOutputwording now match the new line.Not included: the
codex-rescueagent's "If the Bash call fails or Codex cannot be invoked, return nothing" line, which leaves the caller unable to tell a failed run from an empty one. #296 already changes it.Testing
node --test tests/*.test.mjs: 87 pass, 4 fail. The same 4 tests fail on unmodifiedmainin my environment (status shows phases...,status preserves adversarial review kind labels,result returns the stored output...,resolveStateDir uses a temp-backed per-workspace directory), so they are unrelated to this change.tests/commands.test.mjspasses 8/8.