Repository navigation
AGENTS: astra was not short; decide 'short' by dry-run, never from config - #3
Conversation
…run, never from config
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: 46d8b0bf-2ccf-43af-9aa2-00034f18b6f3) |
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. |
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: b8842d50-e845-430a-b302-b661e4bbb0c4) |
Same correction as frdminc/tendcf#29 (see its body).
Review: astra (gpt-6-astra high,
acp-dispatch codex --kind frontier, read-only; admitted at 41% Codex weekly). Verdict: approve with changes. Applied: F1 dry-run readswould refuse:, exits 0 and skips preflight, so the real dispatch must still pass; F2 the estimate is per window from comparable measured history (14 records, scalar and weekly 2 points; the 60% fallback applies per window under three measurements); F3 the claim that astra was available on 10-31 is removed: only that the 45%-vs-60% comparison did not establish shortness.Note
Low Risk
Documentation-only change to review routing guidance; no runtime or policy code is modified.
Overview
Corrects AGENTS.md standing rules so leads do not treat the configured
codex:frontier60% floor—or a raw “% remaining” reading—as proof that astra is unavailable.The doc removes the dated claim that astra was short on 2026-10-31 because Codex was at 45% below that floor. It adds a required procedure: run the real
acp-dispatchinvocation with--dry-run, readwould refuse: class=...(not exit code alone), and remember dry-run skips preflight so a clean dry-run still needs a successful real dispatch. It also clarifies how estimate floors apply per quota window when measured history has fewer than three samples, with a concrete 2026-11-01 history example.Reviewed by Cursor Bugbot for commit b47515e. Bugbot is set up for automated code reviews on this repo. Configure here.