https://github.com/pasunboneleve/skills
Skills for Codex
https://github.com/pasunboneleve/skills
adversarial-review agent-workflows ai-agents beads cheap-to-change codex developer-experience devex llm-guardrails roborev
Last synced: 9 days ago
JSON representation
Skills for Codex
- Host: GitHub
- URL: https://github.com/pasunboneleve/skills
- Owner: pasunboneleve
- License: gpl-3.0
- Created: 2026-05-04T05:29:45.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2026-05-21T06:08:34.000Z (27 days ago)
- Last Synced: 2026-05-21T12:57:06.420Z (27 days ago)
- Topics: adversarial-review, agent-workflows, ai-agents, beads, cheap-to-change, codex, developer-experience, devex, llm-guardrails, roborev
- Language: Shell
- Homepage:
- Size: 1.24 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- License: LICENSE
- Agents: AGENTS.md
Awesome Lists containing this project
README
# [Codex](https://developers.openai.com/codex/cli) skills
[](https://github.com/pasunboneleve/skills/actions/workflows/skill-ci.yml?query=branch%3Amain)
Personal Codex skills used from `~/.codex/skills`. Skill sources live under `src/`.
A sharpened tool. An open codex. A practiced hand.
## Skills
- [`change-friendly-architecture`](src/change-friendly-architecture/SKILL.md): reviews designs for narrow, testable boundaries, explicit dependencies, visible workflows, and low blast radius.
- [`beads`](src/beads/SKILL.md): keeps Beads work claimed, updated with evidence, and closed when complete.
- [`changelog`](src/changelog/SKILL.md): keeps unreleased changelog entries accurate and finalizes version sections at release time.
- [`ci`](src/ci/SKILL.md): adds GitHub Actions checks and README badges for code and skills projects.
- [`coding`](src/coding/SKILL.md): surfaces code execution exceptions, keeps code idiomatic, separates mixed-language scripts, and avoids machine-specific paths.
- [`commit-discipline`](src/commit-discipline/SKILL.md): prepares small local commits with clear rationale and task hygiene.
- [`create-skill`](src/create-skill/SKILL.md): creates concise skills with OpenAI metadata, eval scaffolding, and focused validation.
- [`decompose-skill`](src/decompose-skill/SKILL.md): splits oversized skills into a small trigger surface plus focused resources.
- [`devloop`](src/devloop/SKILL.md): discovers [`devloop`](https://github.com/pasunboneleve/devloop) from its built-in guidance and validates workflows with observable evidence.
- [`documentation-boundary`](src/documentation-boundary/SKILL.md): separates README synopsis work from real documentation updates.
- [`docs-structure`](src/docs-structure/SKILL.md): organises README synopsis content and durable docs into focused files.
- [`kata`](src/kata/SKILL.md): keeps Kata issues searched, assigned, updated with evidence, and closed when verified.
- [`markdown`](src/markdown/SKILL.md): edits Markdown with useful Unicode, including box-drawing directory trees.
- [`release`](src/release/SKILL.md): releases through protected main, PR CI, merged release commits, tags, and release workflows.
- [`repo-policy`](src/repo-policy/SKILL.md): configures GitHub repository policies for protected main, PR merges, CI gates, and artifact releases.
- [`roborev-kata-workflow`](src/roborev-kata-workflow/SKILL.md): coordinates kata task tracking with non-blocking Roborev review.
- [`shell-script`](src/shell-script/SKILL.md): requires strict shell mode for shell scripts and automation wrappers.
- [`skill-project`](src/skill-project/SKILL.md): scaffolds standalone skill-set repositories with shared validation, linking, eval, and agent metadata conventions.
- [`testing`](src/testing/SKILL.md): selects the narrowest deterministic validation for a change's blast radius.
- [`web-realtime-devloop`](src/web-realtime-devloop/SKILL.md): enforces a live browser development loop for realtime web work.
## Use
Run `./scripts/link_skills.sh` to symlink the skills into `~/.codex/skills` and the hosted Codex-home instructions into `~/.codex/AGENTS.md`.
## Validation
Run `bash scripts/validate_skills.sh`. The wrapper installs [`skill-validator`](https://github.com/agent-ecosystem/skill-validator) when needed and uses the verified [`skilpel`](https://github.com/pasunboneleve/skilpel) release binary at `~/.local/bin/skilpel`.
Model-backed evals use [`scripts/skilpel.yaml`](scripts/skilpel.yaml), which sets target and judge temperature to `0`.
To iterate on one eval case, pass a skill relpath and eval id:
```bash
bash scripts/validate_skills.sh --eval-id reject-overfit-domain-type change-friendly-architecture
```
For local model-backed evals, put `OPENAI_API_KEY` in `.env`. The committed `.envrc` loads `.env` into the shell with direnv; `.env` is ignored by Git.
## Versioning
The current version is recorded in `VERSION`.
See `CHANGELOG.md` for release history. Version tags use `vMAJOR.MINOR.PATCH`.
## License
This project is licensed under GPL-3.0-or-later. See `LICENSE`.
