Agent skills extracted from real work. Each one shipped something first.
npx skills add crafter-station/skillsThat lists every skill and lets you pick. To install one directly:
npx skills add crafter-station/skills --skill supply-chain-audit -gWorks with Claude Code, Cursor, Copilot, and 10+ more agents.
Grouped by what the skill does for you. The Status column is explained under Maturity.
Read-only. Produces a verdict, never modifies anything.
| Skill | Version | Status | What it does |
|---|---|---|---|
| supply-chain-audit | 1.0.0 | stable | Scans a developer machine for npm/PyPI supply-chain compromise (Shai-Hulud 2.0, Mini Shai-Hulud / TeamPCP, Axios DPRK). Versioned IOC pack, three-phase scan, PASS/FAIL verdict, and 48h bake-period remediation |
Produces a new artifact from scratch.
| Skill | Version | Status | What it does |
|---|---|---|---|
| generate-brand-assets | 1.0.0 | stable | Generates OG images (1200x630) and favicon formats from a project's branding, with brand colors, gradients, PNG/WebP/ICO |
| surface-recon | 0.6.0 | candidate | Maps what a service, site, desktop app, or file format exposes, and produces a recon report an implementer can build from. Classifies the target into one of eight terrains, each with the technique that works and the gates that stop a confident wrong report |
| cli-build | 0.12.0 | candidate | Builds a CLI an agent can operate and a human can supervise. Distribution target first, then the command surface, polished human output by default, and safety proportional to stakes. Delegates TypeScript primitive adoption to the canonical cligentic skill instead of carrying a copied catalog |
| skill-gen | 0.9.0 | deprecated | Generated skills from documentation. See Maturity before installing |
Orchestrates a release without missing steps.
| Skill | Version | Status | What it does |
|---|---|---|---|
| obsidian-plugin-release | 1.0.0 | stable | Atomic release flow for Obsidian community plugins. Bumps the version across manifest.json, package.json, and versions.json, builds, lints, signs an annotated tag, and triggers a workflow that publishes with build-provenance attestation (SLSA in-toto). Passes the Obsidian Community automated review |
Structures a repository so agents navigate it well.
| Skill | Version | Status | What it does |
|---|---|---|---|
| intent-layer | 1.0.0 | stable | Sets up hierarchical AGENTS.md files so agents navigate your codebase like senior engineers. Built on The Intent Layer by Tyler Brandt |
Category and maturity are separate axes. The category says what a skill does; the maturity says how much evidence stands behind it. A build skill can be stable, and an audit skill can be experimental.
The source of truth is maturity.json, which carries the evidence for each status in one sentence.
| Status | Meaning |
|---|---|
stable |
Used on real work. Recommended |
candidate |
Works, with thin evidence so far |
experimental |
Coherent method and trigger boundary, no real-work use yet |
deprecated |
Kept for provenance. Not recommended |
How a skill moves:
- Nothing enters as stable. Extracted from work actually done starts at
candidate. Written from documentation starts atexperimental. - Promotion needs a named reason, not a feeling. Which work it ran on, and what it caught.
deprecatedis not deletion. The folder stays and the install command keeps resolving. Anyone who already installed it deserves to find out it is unmaintained rather than hit a dead link.- Demotion is normal. A skill that stopped earning its place says so.
Each skill carries a version in its frontmatter, tracking the skill as a document rather than any tool it wraps.
| Bump | Means |
|---|---|
| major | The method changed: a phase added or removed, a reordered flow. Following the old version gives different results |
| minor | Material added, or a claim corrected, without changing the flow |
| patch | Wording, links, typos |
A skill reaches 1.0.0 when it becomes stable: run on real work with the evidence recorded. Candidates stay on 0.x, which is the honest signal that the contract can still move.
A skill that documents one tool belongs with that tool, so it cannot drift from the commands it describes. Those live in the tool's repository:
| Skill | Install |
|---|---|
| skillkit | npx skills add crafter-station/skill-kit |
| spoti-cli | npx skills add crafter-station/spoti-cli |
| sismo-cli | npx skills add crafter-station/sismo-abierto |
| cligentic | bunx --bun skills add Railly/cligentic --skill cligentic |
This repo keeps the skills that are not tied to a single codebase.
skills/<name>/SKILL.md the installable catalog, flat
.claude-plugin/marketplace.json plugin manifest
maturity.json category, status, and evidence per skill
The layout is flat on purpose. The skills CLI discovers skills/<name>/SKILL.md, and skills placed outside that container are invisible to the installer. Measured: a nested skills/<category>/<name>/ layout found 1 of 3 skills in an isolated test, while the flat layout found 3 of 3.
So category is metadata, not a directory. It lives in maturity.json and drives the README grouping, which also means a skill can be recategorized without moving files or breaking a link.
- Fork
- Follow the Agent Skills spec
- Add an entry to
maturity.jsonwith the evidence behind its status - PR
MIT