441 skills across 39 categories. The source of truth for all local skills used by cue profiles.
skills/
├── skills/ The skill library
│ ├── media/ Image and video generation, editing
│ ├── gstack/ Git workflow, reviews, planning, dev utilities
│ ├── meta/ Profile management, skill authoring, memory
│ ├── marketing/ Content, social, campaigns
│ ├── rust/ Rust development, cargo, tooling
│ ├── design/ UI/UX, branding, SVG, Remotion
│ ├── medusa/ Medusa v2 ecommerce
│ ├── nvidia/ GPU, cuOpt optimization
│ ├── content/ Writing, articles, threads
│ ├── research/ Search, papers, keywords
│ ├── review/ Code review, security, testing
│ ├── github/ GitHub CLI, CI fixes, auth
│ └── ... See the full table below
├── scripts/ Install, sync, lint scripts
├── plugins/ Claude Code plugin definitions
└── catalog/ Auto-generated skill index
Each skill is a directory with a SKILL.md file:
---
name: my-skill
description: "Use when the user asks to do X. Does Y."
allowed-tools: Bash(git:*), Read, Write
category: tools
---
# Skill Name
Instructions for the model...Frontmatter rules (enforced by the linter, see below):
name:— required, the discovery name exposed to coding agents.description:— what the LLM matches against to decide when to use the skill. Keep it under 200 characters and include a trigger phrase ("Use when ...").allowed-tools:— useBash(name:*)for shell commands and bare names for top-level tools (Read,Write,Edit, ...). Not["tool_name"].category:/tags:/domain:— at least one, for discoverability.
The repository-relative category/slug path is the canonical library ID. A
name: may remain user-friendly, but duplicate names are reported because a
flat agent runtime cannot load them unambiguously.
Every SKILL.md is checked against the cue spec (rules R001–R013): frontmatter
fields, allowed-tools syntax, trigger phrases, voice rules, anchor links, and
more. Run it locally:
cue skills-lint my-skill # one skill
# or against the whole library from the cue repo:
bun src/index.ts lint-skill skills/CI runs the same checks via skill-md-lint-action. Repository organization is checked separately with:
python scripts/library_inventory.pyThis verifies catalog portability and freshness, generated/template metadata, README counts, canonical IDs, and structural collisions. Existing duplicate names and category mismatches are warnings until their migrations are reviewed; new generated drift and stale artifacts fail CI.
Removed duplicate IDs remain discoverable in
catalog/aliases.json. Consumers should resolve an
alias before looking up the canonical category/slug ID.
Profiles reference skills by category/slug:
# profiles/backend/profile.yaml
skills:
local:
- review/code-review
- deployment/coolify
- github/gh-fix-ciAt launch, cue symlinks these into the runtime's skills/ directory.
cue skills-new my-skill # scaffold
# edit skills/<category>/my-skill/SKILL.md
cue skills-lint my-skill # validate
cue skills-test my-skill # testOr manually: create skills/<category>/<slug>/SKILL.md with the frontmatter format above.
To import an existing user's Codex or Claude Code skills safely, start a new session in this repository with one of the reusable prompts:
The prompts preserve complete skill directories, report semantic collisions instead of overwriting them, and refresh the cue catalog after import.
| Category | Skills |
|---|---|
media |
61 |
gstack |
53 |
meta |
49 |
marketing |
45 |
rust |
41 |
design |
22 |
medusa |
17 |
legal |
14 |
nvidia |
13 |
content |
11 |
research |
10 |
vercel |
9 |
review |
7 |
tools |
7 |
video |
7 |
caveman |
6 |
github |
6 |
orchestration |
6 |
ssh |
6 |
strapi |
6 |
plan |
5 |
higgsfield |
4 |
hostinger |
4 |
obsidian |
4 |
browser |
3 |
career |
3 |
deployment |
3 |
eu-funding |
3 |
security |
3 |
colony |
2 |
polymarket |
2 |
stripe |
2 |
event-design |
1 |
google-workspace |
1 |
predict-everything |
1 |
private |
1 |
robotics |
1 |
test |
1 |
xbot |
1 |
- cue — the profile manager
- resources/mcps — MCP server registry