[Aymericr]

What does this PR do?

Local Claude Code, Codex and other coding agents were required to obtain a named Security or fleet-bot review before completing otherwise authorized work. Replace that rule with the ordinary GitHub workflow: green required CI, any reviews GitHub requires, and resolved P0/P1 findings on the current head. Requests to leave work for review or unmerged remain part of the task scope.

Explicit human authorization still applies to secrets, spend, access changes, sign-ins, destructive production data changes and external messages. Existing authorization does not need to be requested again. This change affects developer instructions only.

How to test

1. Read the working agreement in AGENTS.md: a local coding agent has no named-bot or external-service approval dependency.
2. Confirm it retains required CI, GitHub-required reviews, P0/P1 resolution, task scope and human authorization for the listed production actions.
3. git diff --check passed locally. The instruction audit also confirmed no named-bot approval gate remains in the edited file.

Screenshots / screen recording

N/A — developer instructions only.

Checklist

• I've tested this locally with bun dev
• My code follows the existing code style (run bun check to verify)
• I've updated relevant documentation (if applicable)
• This PR targets the main branch

---

Note

Low Risk
Documentation-only change to agent merge instructions with no runtime or security code impact.

Overview
Updates the Working agreements bullet in AGENTS.md so local agents (Claude Code, Codex, etc.) follow normal GitHub merge rules instead of requiring named Security or fleet-bot reviews.

Merge policy now emphasizes green required CI on the current head, any GitHub-required reviews, resolved P0/P1 comments, and honoring requests to leave work unmerged—explicitly stating that no named bot, bot label, or external approval service is required.

Human gates are reframed: still required for secrets, spend, access, sign-ins/2FA, destructive or access-widening production data changes, and external messages sent under a person's name, with a note not to re-prompt when authorization already exists. CI language is tightened to forbid weakening, skipping, or bypassing checks or branch protection.

^{Reviewed by Cursor Bugbot for commit 976b940. Bugbot is set up for automated code reviews on this repo. Configure here.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T14:36:29.512401Z	debb61c	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.