Skill Slice

Copy one unique skill folder — a slice, not the pie. Do not copy a tag, a parent pack, or the whole tree.

Start with the 12 that were Read-verified (procedure + claimed files, commands, or APIs). That is not a runtime bake. Untested copies are not a claim they work.

docs/WORKS_VERIFIED.md — rate these first.

Browse all unique chairs: Skill Slice explorer (search, filter, copy one install command). Do not --all.

npx skills add kazimrmerchant/skill-slice --skill better

Also: npx skills add kazimrmerchant/skill-slice@better

Swap better for any name on that list. Do not --all. Do not --skill '*'.

Fallback (clone, then one folder):

git clone https://github.com/kazimrmerchant/skill-slice.git skill-slice
Copy-Item -Recurse .\skill-slice\skills\<name> $env:USERPROFILE\.cursor\skills\

Other hosts: copy that same skills/<name>/ folder into the product’s skills root (SKILL.md required). Never install into ~/.cursor/skills-cursor/ (Cursor-managed built-ins).

Progressive-disclosure skills (name + description trigger; body loads when the task matches) for Cursor, Claude, Hermes, and other agentskills.io hosts. Tags (godot, frontend, research, …) are catalog indexes only — not parent directories.

496 unique sibling folders on disk. 12 works-verified. The rest are untested copies — not a claim they work. Read sample was N=24, not a bake of the tree. Method: docs/QUALITY.md. Human browse: explorer.

GitHub has no file-pin API; the repo homepage URL plus docs/WORKS_VERIFIED.md is the pin.

Read AGENTS.md first (30 seconds). Copy one folder. Read that SKILL.md. Do not ingest the tree. Do not start at docs/CATALOG.md.

• Progressive-disclosure skills: name + description are the trigger; the body loads when the task matches.
• A uniqueness bar: independent QUALITY_KEEP, one capability owner, no flood families.
• A works-bar (executability): numbered procedure with real commands / files / APIs — not a stale encyclopedia. See docs/INCLUSION.md. Rate docs/WORKS_VERIFIED.md first.
• MIT for original docs and original skills. Not every body is original — see NOTICE.
• Scorecard dimensions: docs/VALUE_FRAMEWORK.md.

• Private library-only skills
• Quarantine / dual-use offensive content
• Exact body-duplicate losers and flood families (azure-*, godot-genre-*, performing-*, cursor-rules-for-*, wordpress / seismic / pymatgen / book-sft twins)
• Nested mega-packs, node_modules, junctions, skills-inside-skills
• art-direction-islamic-mv (private)
• last30days — install the upstream skill instead (below). We do not vendor that engine.

Full include/exclude: docs/INCLUSION.md.

For “what’s happening now” across Reddit, HN, Polymarket, GitHub, and the web (last 30 days), install the upstream skill — do not expect it in skills/:

npx skills add mvanhorn/last30days-skill

Upstream: mvanhorn/last30days-skill (MIT). Use v3.x. Free sources work without keys. Do not add paid API keys unless you choose to. Complementary to end-to-end-research in this repo (cited due-diligence). Never nest last30days inside another skill.

AGENTS.md · docs/WORKS_VERIFIED.md · docs/QUALITY.md · docs/CATALOG.md (lookup) · docs/INCLUSION.md · NOTICE · CONTRIBUTING.md · SECURITY.md · optional llms.txt

• Review bundled scripts before running them.
• Do not paste API keys into chat or commit .env.
• chrome-browser-automation and grokimagine drive a real Chrome over CDP. They are not anti-bot / CAPTCHA tools.

PII greps and sanitize rules: SECURITY.md. Published files are sanitized; this repo’s git history is not rewritten and is not claimed PII-clean.

MIT for original material. Upstream / community skills: NOTICE.

CONTRIBUTING.md. New skills are sibling folders under skills/<name>/, tagged in the catalog — not nested pack parents.