Repository navigation
AGENTS: standing rule for most-important work (pedantic, Opus lead, cross-vendor frontier review) - #24
Conversation
…ross-vendor frontier review)
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: 64927538-e425-41b8-a85b-4c3d70a459e1) |
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. |
…fined, runnable commands, mechanics not overridden)
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: 301fc71e-6a7e-4f7e-8f73-4f87caffd648) |
Operator standing rule of 2026-10-31 for tendcf and nix2cf (same text in frdminc/nix2cf#1). Memory: site-private memory/feedback_tendcf_nix2cf_pedantic_frontier_review.md.
Review: Fable 5.1 high (
acp-dispatch claude --model fable --kind frontier, read-only), astra short (codex weekly 45% < 60%codex:frontierestimate floor). Verdict: approve with changes. Fixed in the follow-up commit: F1 Argon/acp-run exception, F2 "lifts the reason test, not the mechanics", F3 definition of short, F4 cross-model wording, F5 runnable commands with --perm read-only and --frontier-reason, F6 Argon never counts twice, F8 adversary round is additional, F9 docs-only PRs and handoff exception, F10 nix2cf symlink note, F12 Carbon probe command. F7 and F11 confirmed by djbclark: Opus for every lead session including pipeline-only; code slices Sonnet 5.5 high or better, never a weaker tier. F13: the GitHub codex-connector app review is not the frontier review.Note
Low Risk
Documentation-only policy for AI agents; no runtime, auth, or deployment behavior changes.
Overview
AGENTS.md now codifies operator standing rules (2026-10-31) for tendcf and nix2cf: ordinary changes must go through a PR with cross-model frontier review, while
docs/handoffs/keeps the push-in-place exception.The new section replaces vague “normal review as directed” with five concrete policies: pedantic verification (run checks, don’t claim verified without running), model tiers (Opus 5.5 high for every lead including pipeline-only; code slices Sonnet 5.5 high or better via
/bigteam; no weak tiers), and mandatory cross-vendor frontier review before merge (astra → Fable → Gemini Argon, with “short pool” fallbacks, two reviews for design rulings, PR-body documentation, and explicit bans on same-family or GitHub-app reviews). It also states thatmodel-routing1b’s reason test is lifted for these repos only—notacp-dispatchquota mechanics—and adds a weekly probe note for Gemini 4 Carbon when it becomes reachable.Reviewed by Cursor Bugbot for commit cd7cb26. Bugbot is set up for automated code reviews on this repo. Configure here.