Codex in VS Code Template

# AGENTS.md — Project Instructions

You are the Codex IDE extension running in VS Code. Read this file before any work. Keep it under ~200 lines — push deep workflows into Skills under `.agents/skills/`.

This file is shared with the Codex CLI and Codex App via the same `~/.codex/` config — anything you write here applies in all three surfaces.

## Discovery (how Codex layers context)
1. **Global:** `~/.codex/AGENTS.override.md` (if present), else `~/.codex/AGENTS.md`
2. **Project:** repo root → CWD, walking down. Each directory: `AGENTS.override.md` → `AGENTS.md` → fallback names
3. **Merge:** files closer to your CWD override earlier guidance
4. **Cap:** combined size limited by `project_doc_max_bytes` (32 KiB default)

## Meta-Learning Protocol

You operate under The Learner Brain methodology. Improve the *system*, not just the output.

### Upgrade Triggers
1. **Repetition** — same correction twice
2. **Friction** — workaround the user shouldn't repeat
3. **Manual labor** — multi-step task that should become a Skill
4. **Preference** — codify expressed style/approach

### Upgrade Format
```
🧠 BRAIN UPGRADE PROPOSAL
Trigger: [what caused this]
Surface: [AGENTS.md | Skill | .vscode/settings.json]
Location: [path]
Proposed change: [exact addition]
Rationale: [why it prevents future friction]
```

### Hygiene
- AGENTS.md cap: ~200 lines per file. Beyond that, split into a nested AGENTS.md or a Skill.
- Never edit AGENTS.md without approval. Propose ONE upgrade at a time.

---

## VS Code Workflow

- Use the **Codex panel** (left sidebar icon) for chat + agent runs.
- Inline edits: select code → Codex action → review diff → accept/reject in the editor gutter.
- Approval mode: prefer **suggest** for risky tasks, **auto-edit** only for trusted Skills.
- Reference files in chat with `@filename`. Reference terminal output with `@terminal`.

---

## Project Context

### Tech Stack
- Language: [Your language]
- Framework: [Your framework]
- Package manager: pnpm (preferred)
- Test: [test command]

### Working Agreements
- Run lint + tests before claiming a task is done
- Ask before adding production dependencies
- Never commit `.env*` or anything under `secrets/`

## Patterns Learned
<!-- Append discovered patterns. Promote to Skills when they grow. -->
- [Pattern]: [solution]

## Anti-Patterns
- Do not regenerate boilerplate when an existing Skill covers it
- Do not push directly to main

// .vscode/settings.json — workspace defaults for the Codex extension
{
  // Default approval mode for this workspace.
  // "suggest" = show diffs, never auto-apply.
  // "auto-edit" = apply file edits without asking, still confirm shell.
  // "full-auto" = no prompts (use only with sandboxing).
  "codex.approvalMode": "suggest",

  // Preferred model + reasoning effort for this project.
  // Current flagship: gpt-5.5. Fallbacks: gpt-5.4, gpt-5.4-mini, gpt-5.3-codex.
  "codex.model": "gpt-5.5",
  "codex.reasoningEffort": "medium",

  // Restrict the agent's filesystem + network surface.
  "codex.sandbox": "workspace-write",

  // Auto-load AGENTS.md from project root + ~/.codex (default true; explicit here).
  "codex.loadAgentsMd": true,

  // Surface the Codex panel on startup.
  "codex.openOnStartup": false
}

# ~/.codex/config.toml — shared by VS Code extension, CLI, and App

# Raise the AGENTS.md byte cap (default 32 KiB)
project_doc_max_bytes = 65536

# Treat additional filenames as instruction files
project_doc_fallback_filenames = ["TEAM_GUIDE.md", ".agents.md"]

[sandbox]
# read-only | workspace-write | danger-full-access
mode = "workspace-write"

# [[skills.config]]
# path = "/absolute/path/to/.agents/skills/my-skill/SKILL.md"
# enabled = true