
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
This repo packages 35 reusable skills that agents can load as plain `SKILL.md` folders. The same skill content works across Codex, Hermes, and Claude; only the install location changes. It also includes Vietnamese documentation, provenance files, checksum manifests, and scripts to build and verify the bundle.
Builders who want to add reusable skill packs to Codex, Claude, or Hermes sessions.
You can install the same skill set across agent runtimes instead of rebuilding prompts and workflows for each one.
Skills cover writing, research, data, documents, presentations, design, media, finance, deployment, and skill creation.
Each skill lives in its own `SKILL.md` folder, with extra files only when needed.
`packages/` contains one ZIP per skill for Claude skill-management flows.
PowerShell and shell installers place skills into the right directory for Codex, Hermes, or Claude.
`verify_bundle.py` and the checksum files check structure, provenance, and archive sync.
`GIAI_THICH_SKILLS_VI.md` explains every skill in Vietnamese with purpose, inputs, outputs, and examples.
.\install-codex.ps1
bash install-hermes.sh
bash install-claude.sh
.\install-claude.ps1
python build_claude_packages.py --clean
Codex: ~/.agents/skills/<skill-name>
Hermes: ~/.hermes/skills/<skill-name>
Claude: ~/.claude/skills/<skill-name> personal, every session
<project>/.claude/skills/<skill-name> one project onlyA personal collection of portable Agent Skills curated, adapted, and maintained by Rylai for Codex, Hermes, and Claude.
The bundle contains 35 independent skills for writing, research, data, documents, presentations, design, media, finance, deployment, and skill creation. The skills can be installed together or copied individually.
Every skill is a plain SKILL.md folder with no runtime lock-in, so the same
files load in Codex, Hermes, and Claude. Only the install location differs.
See THIRD_PARTY_NOTICES.md and PROVENANCE.yml for the release provenance.
See GIAI_THICH_SKILLS_VI.md for a Vietnamese description of every skill, including its purpose, expected inputs and outputs, runtime requirements, and example requests.
.\install-codex.ps1
Use -Force to replace existing skills. Existing folders are backed up first.
bash install-hermes.sh
Use --force to replace existing skills. Existing folders are backed up first.
bash install-claude.sh
.\install-claude.ps1
Use --force (-Force in PowerShell) to replace existing skills. Existing
folders are backed up first.
The default target is the personal skill directory, so the skills load in every
Claude session. Pass --project (-Project) to install into .claude/skills
next to the current project instead, or --target <dir> (-Target <dir>) to
choose the directory yourself. Skills are discovered when a session starts, so
start a new Claude session after installing.
Claude reads the same SKILL.md frontmatter as the other runtimes. The extra
agents/openai.yaml file in each skill is ignored there and is kept only so one
checkout serves all three agents.
packages/ holds one .zip archive per skill, built from skills/ by
build_claude_packages.py. Each ZIP has the skill folder at its root, including
that folder's SKILL.md and bundled resources.
Use these with Claude.ai or Cowork through the skill-management interface.
Claude Code should use the unpacked folders through install-claude.sh,
install-claude.ps1, or a manual copy into a Claude skill directory.
Rebuild them after editing any skill:
python build_claude_packages.py --clean
Archives are reproducible, so an unchanged skill keeps its checksum. Per-archive
hashes are written to packages/checksums.sha256.
Copy one directory from skills/ into the skill directory used by your agent:
Codex: ~/.agents/skills/<skill-name>
Hermes: ~/.hermes/skills/<skill-name>
Claude: ~/.claude/skills/<skill-name> personal, every session
<project>/.claude/skills/<skill-name> one project only
ready: instruction workflow is usable without a special proprietary runtime.conditional: requires an optional package, binary, login, API, or renderer.adapted-core: the portable core is available; advanced integrations may be
intentionally omitted.Runtime requirements are described in each SKILL.md and in manifest.json.
python -m pip install -r requirements-verify.txt
python verify_bundle.py
The verifier checks skill structure, UI metadata, Rylai maintainer metadata,
provenance, manifest consistency, Markdown fences and inline code spans, private
paths, banned vendor residue, third-party notices, and whether all 35 Claude ZIP
packages are still in sync with skills/.
checksums.sha256 covers the source files in this repository. The generated
archives are covered separately by packages/checksums.sha256. Both are plain
sha256 manifests:
sha256sum -c checksums.sha256
cd packages && sha256sum -c checksums.sha256
The archive manifest lists bare file names, so check it from inside packages/.
Get-FileHash -Algorithm SHA256 <file>
Regenerate the release checksum file after changing any source file:
python build_claude_packages.py --release-checksums
skills/
<skill-name>/
SKILL.md
agents/openai.yaml
scripts/ references/ assets/ ... when required
manifest.json
PROVENANCE.yml
THIRD_PARTY_NOTICES.md
packages/
<skill-name>.zip
checksums.sha256
GIAI_THICH_SKILLS_VI.md
install-codex.ps1
install-hermes.sh
install-claude.ps1
install-claude.sh
build_claude_packages.py
verify_bundle.py
Rylai-authored material is released under the MIT License. Third-party material keeps its original license; see the notices and provenance files.
This is a community project and is not endorsed by the agent platforms or the upstream repositories listed in the notices.
Sign in to join the discussion.
No comments yet. Be the first to say what this is good for.

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)
A skill to stop your coding agent from burying the answer. ADHD-friendly output.
AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture repair for fiction, venue-matched rules for professional prose. Based on StoryScope (arXiv:2604.03136).
Academic Research Skills for Claude Code: research → write → review → revise → finalize