codex

<!-- GENERATED FILE -- do not edit by hand.

This file is regenerated from AGENTS.md by scripts/generate_agent_configs.py. Bootstrap re-runs the generator on every session, so edits here are lost.

Precedence for agent rule files (most specific wins): 1. agents/codex.local.md your per-agent, per-project overrides 2. AGENTS.local.md your cross-agent, per-project overrides 3. agents/codex.md generated from AGENTS.md (this file) 4. AGENTS.md upstream baseline

To customize just for Codex in this project, create agents/codex.local.md (the generator never touches it). To customize for every agent in this project, edit AGENTS.local.md. To change upstream rules for everyone, edit AGENTS.md in your fork. -->

Shared agent rules (anywhere-agents)

This file holds rules. A rule says what to do, plus one clause on why when the rule is counter-intuitive. Longer rationale, measurements, version history, and how-to material go to the docs linked at the end or to `CHANGELOG.md`, never here. Tests enforce the size of this file.

<!-- Byte-identical in anywhere-agents and agent-config (check-parity STRICT); size gate in tests/test_bootstrap_size.py; maintainer-only lines go in AGENTS.local.md. -->

> **Determine your context before proceeding.** > > **Dispatched reviewer:** A dispatched reviewer whose task, lens, and response format are supplied is exempt from this file's bootstrap, router, and coordinator-skill discovery instructions; it applies substantive project policy, including applicable local overrides, reads missing instruction files once, and may read any of these files when they are themselves in scope or answer a concrete verification question. > > **Source repo test:** If `bootstrap/bootstrap.sh`, `bootstrap/bootstrap.ps1`, `scripts/generate_agent_configs.py`, and `skills/` all exist at the repo root, you are in a source repo (`anywhere-agents`, `agent-config`, or a fork with the same layout). Skip the bootstrap block below and proceed directly to `## Session Start Check`. Do not execute the bootstrap script in a source repo; it would try to bootstrap itself. > > **Consumer repo path:** Otherwise, you are in a project that consumes the shared config. **You MUST execute the matching bootstrap block below at the start of every session, before responding to any user request or running session checks.** Use the PowerShell block on Windows and the Bash block on macOS/Linux. Bootstrap is idempotent: run it every session even if `.agent-config/` already exists, so shared config, skills, and settings are refreshed from upstream before any work begins. After running, report the refresh result in one line (for example, "bootstrap: refreshed AGENTS.md, skills, settings").

Bootstrap

Project repos carry this block at the top of their `AGENTS.md`; bootstrap rewrites it there on every run, so it is both the reference and the runtime command.

## Shared Agent Config (auto-fetched)

Before starting any task, fetch the shared agent configuration and skills. Use the block for your shell.

PowerShell (Windows):

```powershell
New-Item -ItemType Directory -Force -Path .agent-config, .claude, .claude/commands | Out-Null
Invoke-WebRequest -UseBasicParsing -Uri https://raw.githubusercontent.com/yzhao062/anywhere-agents/main/bootstrap/bootstrap.ps1 -OutFile .agent-config/bootstrap.ps1
& .\.agent-config\bootstrap.ps1

Bash (macOS/Linux):

mkdir -p .agent-config .claude/commands
curl -sfL https://raw.githubusercontent.com/yzhao062/anywhere-agents/main/bootstrap/bootstrap.sh -o .agent-config/bootstrap.sh
bash .agent-config/bootstrap.sh

Read and follow the rules in `.agent-config/AGENTS.md` as baseline defaults. Any rule in `AGENTS.local.md` overrides the shared default.

Rules that follow from the run:

- Bootstrap rewrites the consuming repo's root `AGENTS.md` on every run, composing the shared copy with the configured passive packs (by default `agent-style`, which needs Python 3 and PyYAML), then regenerates `CLAUDE.md` and `agents/codex.md` from it. When packs are configured but composition cannot run, an existing composed `AGENTS.md` is preserved rather than replaced by the bare shared copy; that run warns and records `completed: false` in `.agent-config/last-run.json`.
- Put repo-local overrides in `AGENTS.local.md`, never in the generated files. Read `AGENTS.local.md` after `AGENTS.md`; its rules win. Codex does not discover it on its own, which is why this sentence exists.
- Shared command pointers are copied into `.claude/commands/`; the copy overwrites same-named files only and does not delete unrelated project-local commands.
- Shared keys of `.claude/settings.json` (`permissions`, `attribution`, `effortLevel`, and the like) are merged on every run and project-only keys are preserved. Override a shared key locally in `.claude/settings.local.json`.
- Bootstrap also refreshes user-level files: `~/.claude/hooks/guard.py` (PreToolUse), `~/.claude/hooks/session_bootstrap.py` (SessionStart), the statusline scripts, and the shared `env` entries of `~/.claude/settings.json`.
- `.agent-config/`, `agent-config.local.yaml`, and the three generated files (`AGENTS.md`, `CLAUDE.md`, `agents/codex.md`) are gitignored in a consumer; a generated file is never untracked automatically once a repo tracks it; moving one out of the index is an operator decision. `agent-config.yaml` selects packs under the `packs:` key (`rule_packs:` is deprecated). `todo/` holds files a person drops in for an agent; what an agent generates goes to the session scratchpad, not there.

### Configuration Precedence

Rule files, most specific wins:

| Layer | File | Scope |
|---|---|---|
| 1 | `CLAUDE.local.md` / `agents/codex.local.md` | Per-agent, project-local, hand-authored; bootstrap never touches it |
| 2 | `AGENTS.local.md` | Cross-agent, project-local, hand-authored; bootstrap never touches it |
| 3 | `CLAUDE.md` / `agents/codex.md` | Per-agent, generated from `AGENTS.md` by `script