Repository navigation
docs(implementation): qualify model and effort pin claims - #11573
Conversation
The dispatch and agent text said a frontmatter pin always decides the tier. Upstream lists sources ranked above it, so point at those sections with an as-of date and recheck trigger. Refs: #6916 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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. |
|
claude-security-review reviewed No security findings. The diff is four prose-only files: a changelog fragment and the bodies of Instruction-surface check. Two edits soften wording, so I read them as possible rule removals.
Neither edit removes a control. Both correct an overclaim by saying the frontmatter pin can be outranked upstream. Nothing was removed that a hook, permission rule or validator would otherwise have to enforce. I did not consult Other changes. The Spend. Nothing in the diff adds runs, turns, models or schedules, so there is no spend to report. critical-findings: 0 |
|
claude-review reviewed No findings. The PR changes documentation only: it qualifies the "pin always holds" wording in
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c5e142d901
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".
|
The pr-verify verifier failed head What it wrote (23 lines)This pull request leaves the merge queue, and ci-status holds it out while it carries |
The dispatch line paraphrased how the harness ranks the binding, so state only this repository's decision and leave the rest to the pointer. Point the effort record at the section heading a raw read can find. Refs: #6916 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Refs: #6916
Summary
Delivers the
plugins/implementationpart of the repo-mode/multi-agent:audit-defaultssweep. The text said a frontmatter model or effort pin always holds; upstream ranks other sources above it.Fix
implement-dispatchno longer says a dispatch lands on the intended tier "regardless": it states this repository's decision and leaves how the binding resolves to two sub-agents sections.scoped-implementerno longer says the phase "never" runs at the session's level, and its record covers every source ranked over the pin.implementeradds amodel-config#model-aliasespointer to its model-binding record.The pilot's fourth finding here, a dead anchor at
scoped-implementer.md:157, was refuted: the raw re-check on 2026-10-31 foundid="recommended-effort-levels-for-claude-sonnet-5-5"still on the rendered page. The link now uses the section heading anyway, because a raw read cannot find that legacy id.No routing or dispatch rule changes. These edits add qualifiers and pointers and cut nothing, so no eval evidence applies.
Verification
sub-agentsandmodel-configpages on 2026-10-31.scripts/check-affected.sh --tests: passed on the new head.claude plugin eval, because this run spawns none and no edit is a cut.Related
Remaining
/multi-agent:audit-defaults repoover every file in its target list, with the count comment on docs(multi-agent,session-flow): fix repo-mode audit pilot findings and schedule a full sweep #6916.ci-statusgreen on each sweep PR.🤖 Generated with Claude Code