Catalog
August 11, 2026 · View on GitHub
Tutorials
| Module | Core artifact |
|---|---|
| 00 · Mental model | Surface and task selection |
| 01 · Sandbox and approvals | Risk-based permission profiles |
| 02 · CLI and surfaces | Safe first-session checklist |
| 03 · Prompts and plans | Four-part prompt contract |
| 04 · AGENTS.md | Root and nested instruction strategy |
| 05 · Engineering skills | Skill anatomy and installation |
| 06 · MCP and tools | Tool selection and safety review |
| 07 · Testing and review | Verify-and-review loop |
| 08 · Subagents | Bounded delegation |
| 09 · Orchestration | Parallel work protocol |
| 10 · Context efficiency | Context budget and summary contracts |
| 11 · Automation and plugins | Promotion path from prompt to automation |
| 12 · Troubleshooting | Layer-by-layer diagnosis |
| 13 · Living Codex wiki | Review-first knowledge compilation |
Installable skills
| Skill | Bundled reference |
|---|---|
| engineering-loop | references/loop-contract.md, references/hard-debugging.md |
| build-frontend | references/verification.md |
| build-backend | references/contracts-and-data.md |
| operate-devops | references/change-safety.md |
| review-code | references/review-checklist.md, references/review-lenses.md |
| test-software | references/test-strategy.md |
| review-security | references/threat-checklist.md |
| orchestrate-engineering | references/context-budget.md |
| maintain-codex-wiki | references/source-policy.md, references/article-template.md, scripts/wiki_lint.py |
Living knowledge
- Codex Living Wiki — review model and operating boundaries.
- Wiki index — compact routing across current topics, decisions, and experiments.
- Source registry — provenance without committing external source bodies.
- Wiki efficiency baseline — planned no-wiki/query/maintenance comparison.
Copy-ready examples
- Starter
AGENTS.md - Conservative project configuration
- Explorer agent
- Reviewer agent
- Test runner agent
- General task prompt
- Orchestrated review prompt
- Scheduled documentation-drift prompt
- Engineering specification
- Dependency-aware ticket set
- Engineering handoff
- Measured task receipt
- Orchestration run receipt
- Edition manifest
- Engineering-loop run record
- Incident orchestration smoke run
- Local plugin marketplace
- Stop-hook example
- Explicit-only workflow router
- Orchestration decision matrix
- Token-efficiency scorecard
- Model-adaptive skill guide
- Third-party workflow calibration
- Community launch kit
- Fork-an-edition guide
- Verified distribution shortlist
- v0.3.0 release notes
- v0.2.0 release notes
Hands-on lab
- Engineering playground — a dependency-free frontend, backend, and deployment fixture with seeded defects for the end-to-end loop and its specialist workflows.
- Incident-response orchestration benchmark — a partitionable full-stack task for controlled single-agent versus orchestrated implementation runs.
- Developer presentation guide — a 15-minute talk track, live-demo script, measurement plan, and copy-ready announcement.
- Engineering-loop measurement protocol — three-way skill ablation, quality gates, metrics, and decision rules.
- Community guide — responsible participation, useful fork paths, and community standards.