Codex Skills

Canonical source repository for reusable Codex/OpenAI workflow skills maintained by KikuAI Lab.

Each top-level directory is one installable skill entry point with its own SKILL.md, agent metadata, and any references, scripts, assets, runtime code, or tests it needs. Related skills may call one another, but products and libraries with an independent audience or release lifecycle stay in their own repositories.

Skill catalog

Planning, product, execution, and review

• work-shaping
• idea-validation
• product-shaping
• spec-bundle
• codex-execution-pack
• continuity-ledger
• issue-control-loop
• adversarial-review
• adr-log
• github-mobile-ops
• justdoit
• lazy-senior
• long-context-dispatch
• parallel-worktrees
• product-council
• triage-finding
• tool-scout

Documentation, communication, and public artifacts

• ai-writing-detox
• illustration-prompt
• public-artifact-lane
• readme-generator
• github-seo
• session-to-post
• visual-explainer

Browser, media, and external workflows

• autonomous-video-pipeline
• browser-tutorial-video
• openclaw-council
• playwright
• screenshot
• video-builder
• video-spec-builder
• video-render-run
• transcribe
• transcribe-basic
• transcribe-diarize

Spreadsheet workflows

• spreadsheet
• spreadsheet-xlsx-edit
• spreadsheet-tabular-analysis

System skills

• .system/skill-creator
• .system/skill-installer

Using a skill

Copy or install the required directory as a complete unit. Do not copy only SKILL.md when the directory also contains agents/, references/, scripts/, assets/, runtime code, or tests.

The video, spreadsheet, and transcription child skills use their corresponding parent directory for shared scripts or references. Install the whole family when using a child entry point.

The public descriptions, use cases, and attribution index live in the repository catalog. Its machine-readable form is catalog/skills.json.

Repository policy

A workflow belongs here when its primary lifecycle is that of a reusable skill for Codex or a compatible agent surface. A standalone product, reusable library, service, operational source of truth, or experiment stays independent even if it also exposes a skill-shaped entry point.

Imported repositories retain their source commits through non-squashed history merges. Superseded source repositories were removed after their canonical copies and histories were verified; former names and destinations are recorded in MIGRATIONS.md.

Validation

Validate all skill metadata and run the imported runtime tests:

python3 -m pip install pyyaml pytest

find . -mindepth 2 -maxdepth 2 -name SKILL.md -print \
  | sed 's#/SKILL.md##' \
  | while read -r skill; do
      python3 .system/skill-creator/scripts/quick_validate.py "$skill"
    done
python3 .system/skill-creator/scripts/quick_validate.py .system/skill-creator
python3 .system/skill-creator/scripts/quick_validate.py .system/skill-installer
python3 -m json.tool catalog/skills.json >/dev/null
python3 scripts/check_public_privacy.py

PYTHONPATH=session-to-post/src python3 -m pytest -q session-to-post/tests
PYTHONPATH=issue-control-loop/src python3 -m pytest -q issue-control-loop/tests
python3 -m pytest -q codex-execution-pack/tests

GitHub Actions runs the same checks on pushes and pull requests.

Contributing and security

See CONTRIBUTING.md for repository boundaries and test expectations. Report vulnerabilities through GitHub's private vulnerability reporting flow when available; see SECURITY.md.

Follow the work

Skill releases and project notes: Telegram · LinkedIn · KikuAI

Licenses

This repository is not uniformly licensed. Original KikuAI Lab material is MIT unless a directory says otherwise. Imported AGPL-3.0 and Apache-2.0 skills keep their own license text. See LICENSES.md before redistributing a skill.

_{Follow new projects and updates from @kiku-jw.}