opencue · NagyVikt · Jun 20, 2026 · Jun 20, 2026 · Jun 20, 2026
diff --git a/README.md b/README.md
@@ -1,30 +1,25 @@
 # cue/skills — Skill Library
 
-> 127 skills across 21 categories. The source of truth for all local skills used by [cue](https://github.com/opencue/claude-code-skills) profiles.
+> 427 skills across 36 categories. The source of truth for all local skills used by [cue](https://github.com/opencue/claude-code-skills) profiles.
 
 ## What's here
 
 ```
 skills/
 ├── skills/                 The skill library
-│   ├── browser/            Playwright, screenshots
-│   ├── caveman/            Terse mode, commits, compression
-│   ├── colony/             Multi-agent coordination
-│   ├── deployment/         Coolify, Supabase, pnpm
+│   ├── 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
-│   ├── github/             GitHub CLI, CI fixes, auth
-│   ├── higgsfield/         AI image/video generation
-│   ├── hostinger/          DNS, domains, VPS
 │   ├── medusa/             Medusa v2 ecommerce
-│   ├── meta/               Profile management, memory, helpers
-│   ├── nvidia/             cuOpt, GPU optimization
-│   ├── obsidian/           Vault, markdown, canvas
-│   ├── orchestration/      Fleet, pipelines, workers
+│   ├── nvidia/             GPU, cuOpt optimization
+│   ├── content/            Writing, articles, threads
 │   ├── research/           Search, papers, keywords
 │   ├── review/             Code review, security, testing
-│   ├── secrets/            Envoult, credential management
-│   ├── stripe/             Payments, webhooks
-│   └── ...
+│   ├── 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
@@ -36,16 +31,39 @@ Each skill is a directory with a `SKILL.md` file:
 
 ```markdown
 ---
-description: "When user says X, do Y"
-allowed-tools: ["tool_name"]
+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...
 ```
 
-The `description` frontmatter is what the LLM matches against to decide when to use the skill.
+Frontmatter rules (enforced by the linter, see below):
+
+- `name:` — required, the canonical id Claude's skill discovery uses.
+- `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:` — use `Bash(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.
+
+## Linting
+
+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:
+
+```bash
+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](https://github.com/opencue/claude-code-skills/tree/main/skill-md-lint-action).
 
 ## How cue uses this
 
@@ -76,15 +94,43 @@ Or manually: create `skills/<category>/<slug>/SKILL.md` with the frontmatter for
 ## Categories
 
 | Category | Skills | Domain |
-|----------|--------|--------|
-| `browser` | 1 | Playwright, screenshots |
-| `caveman` | 5 | Terse mode, commits |
-| `design` | 16 | UI/UX, branding, SVG |
-| `medusa` | 14 | Ecommerce platform |
-| `meta` | 18 | Profile helpers, memory |
-| `nvidia` | 12 | GPU optimization |
-| `review` | 5 | Code quality, security |
-| `research` | 8 | Search, papers |
+|----------|-------:|--------|
+| `media` | 61 | Image and video generation, editing |
+| `gstack` | 53 | Git workflow, reviews, planning, dev utilities |
+| `meta` | 49 | Profile management, skill authoring, memory |
+| `marketing` | 45 | Content, social, campaigns |
+| `rust` | 41 | Rust development, cargo, tooling |
+| `design` | 33 | UI/UX, branding, SVG, Remotion |
+| `medusa` | 17 | Medusa v2 ecommerce |
+| `nvidia` | 13 | GPU, cuOpt optimization |
+| `content` | 11 | Writing, articles, threads |
+| `research` | 10 | Search, papers, keywords |
+| `video` | 7 | Video production flows |
+| `tools` | 7 | Utilities, token/cost analysis |
+| `strapi` | 6 | Strapi v5 CMS |
+| `ssh` | 6 | Remote servers, SSH ops |
+| `review` | 6 | Code review, security, testing |
+| `orchestration` | 6 | Multi-agent, fleets, pipelines |
+| `github` | 6 | GitHub CLI, CI fixes, auth |
+| `caveman` | 6 | Terse mode, commits, compression |
+| `plan` | 5 | Planning, investigation, reviews |
+| `obsidian` | 4 | Vault, markdown, canvas |
+| `hostinger` | 4 | DNS, domains, VPS |
+| `higgsfield` | 4 | AI image/video generation |
+| `security` | 3 | Audits, secrets, hardening |
+| `eu-funding` | 3 | EU grant applications |
+| `deployment` | 3 | Coolify, Supabase, pnpm |
+| `career` | 3 | Resumes, interviews |
+| `browser` | 3 | Playwright, screenshots |
+| `stripe` | 2 | Payments, webhooks |
+| `polymarket` | 2 | Prediction markets |
+| `colony` | 2 | Multi-agent coordination |
+| `xbot` | 1 | X/Twitter automation |
+| `test` | 1 | Edge-case and test generation |
+| `private` | 1 | Private/internal |
+| `predict-everything` | 1 | Forecasting, simulation |
+| `google-workspace` | 1 | Docs, Sheets, Gmail |
+| `event-design` | 1 | Invitations, cards |
 
 ## Related
 

diff --git a/skills/browser/agent-browser/SKILL.md b/skills/browser/agent-browser/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: agent-browser
-description: Use when you need to drive a real browser: open a page, fill forms, click, screenshot, scrape, or test a login flow. Triggers on "automate the browser", "test the page", "scrape this site".
+description: "Use when you need to drive a real browser: open a page, fill forms, click, screenshot, scrape, or test a login flow. Triggers on \"automate the browser\", \"test the page\", \"scrape this site\"."
 allowed-tools: Bash(agent-browser:*)
 category: browser
 tags: [browser, automation, scraping, testing, cdp]

diff --git a/skills/browser/lightpanda/SKILL.md b/skills/browser/lightpanda/SKILL.md
@@ -5,6 +5,7 @@ metadata:
   category: browser
   domain: browser-automation
   tags: [browser, headless, cdp, scraping, mcp, lightpanda]
+category: browser
 ---
 
 # Lightpanda: fast headless browser for agents

diff --git a/skills/browser/playwright/SKILL.md b/skills/browser/playwright/SKILL.md
@@ -1,18 +1,19 @@
 ---
 name: playwright
-description: Use when user says "does it work?", "screenshot the page", "show me the form", "check the checkout flow", or "open the browser". Drives Chromium/Firefox/WebKit via Playwright MCP to verify storefront/admin UI changes, reproduce reported UI bugs, test Medusa dev servers on localhost, or explore third-party sites.
+description: Use when user says "does it work?", "screenshot the page", "check the checkout flow", or "open the browser". Drives a browser via Playwright MCP to verify UI changes or reproduce bugs.
+category: browser
 ---
 
-# Playwright MCP — driving a real browser
+# Playwright MCP, driving a real browser
 
 ## Core principle
 
-When code touches anything rendered (Medusa storefronts, admin dashboards, Coolify UI, third-party docs), running tests or reading source isn't enough — load the page in a real browser, see what users see. Playwright MCP exposes `navigate`, `screenshot`, `click`, `fill`, `evaluate`, `accessibility_snapshot`, `network_log`, and friends so Claude can drive Chromium, Firefox, or WebKit from inside a conversation.
+When code touches anything rendered (Medusa storefronts, admin dashboards, Coolify UI, third-party docs), running tests or reading source isn't enough, load the page in a real browser, see what users see. Playwright MCP exposes `navigate`, `screenshot`, `click`, `fill`, `evaluate`, `accessibility_snapshot`, `network_log`, and friends so Claude can drive Chromium, Firefox, or WebKit from inside a conversation.
 
 ## When to use
 
 - "Does the storefront still build after my change?" → navigate to the local dev URL, screenshot above-the-fold, check console errors.
-- "I changed the checkout button text — confirm it shows up." → navigate, screenshot the cart, verify the text in the accessibility snapshot.
+- "I changed the checkout button text, confirm it shows up." → navigate, screenshot the cart, verify the text in the accessibility snapshot.
 - "Reproduce the 500 on /blog/some-post" → navigate, capture the network_log, surface the failing request.
 - "What does marvahome.com look like in mobile viewport?" → navigate with viewport override, screenshot.
 - "Inspect the DOM of the Stripe Checkout page after redirect" → drive through the checkout, evaluate JS to dump the DOM.
@@ -116,15 +117,15 @@ Browsers live under `~/.cache/ms-playwright/`.
 
 | Symptom | Cause | Fix |
 |---|---|---|
-| `Target page, context or browser has been closed` | session timed out between calls | `navigate` again — Playwright MCP creates a new context per task |
+| `Target page, context or browser has been closed` | session timed out between calls | `navigate` again, Playwright MCP creates a new context per task |
 | Screenshots are blank / white | page hasn't finished rendering | `wait_for({ load_state: "networkidle" })` first |
 | Selector matches nothing | shadow DOM or iframe | use `accessibility_snapshot` to find the right role/name; for iframes pass `frame: "name"` |
 | Slow first navigate | Chromium cold-start (~2s) | normal; subsequent calls reuse the browser |
 | Network log empty | `network_log` only captures since the last `navigate` | call after the page settles |
 
 ## When to escalate to Chrome DevTools MCP
 
-If you need to interact with your *already-open*, *already-logged-in* Chrome session (e.g., the Coolify dashboard with your live token), Playwright won't help — it always starts a fresh incognito-style context. For that, install `@modelcontextprotocol/server-chrome-devtools`, launch Chrome with `--remote-debugging-port=9222`, and use that MCP. Both can coexist.
+If you need to interact with your *already-open*, *already-logged-in* Chrome session (e.g., the Coolify dashboard with your live token), Playwright won't help, it always starts a fresh incognito-style context. For that, install `@modelcontextprotocol/server-chrome-devtools`, launch Chrome with `--remote-debugging-port=9222`, and use that MCP. Both can coexist.
 
 ## When to escalate to Computer Use
 

diff --git a/skills/career/job-hunter/SKILL.md b/skills/career/job-hunter/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: job-hunter
-description: Run a whole job-search campaign, not a one-off document. Use when the user says "find me jobs", "run my job search", "job hunt", "am I competitive for this role", "build my application package", "prep me for interviews", or "negotiate this offer". Routes live search, fit-scoring, ATS resume and cover letter, interview prep, and offer negotiation across four stages. For a single resume rewrite use resume-tailor; for JD scoring use jdfit.
+description: Use when the user says "find me jobs", "run my job search", "job hunt", or "prep me for interviews". Routes live search, fit-scoring, ATS resume, interview prep, and offer negotiation.
 allowed-tools: Read, Write, Glob, Grep, WebSearch, WebFetch, Bash(docker:*)
 argument-hint: [stage or goal, e.g. "search recruiter roles Budapest" or "tailor for <JD url>"]
 tags: [career, job-search, resume, interview, recruiting]

diff --git a/skills/career/resume-version-manager/SKILL.md b/skills/career/resume-version-manager/SKILL.md
@@ -1,6 +1,7 @@
 ---
 name: resume-version-manager
-description: Track resume versions, maintain one master resume, and manage per-job tailored copies. Use when user says "manage my resumes", "track resume versions", "which resume did I send", "master resume", "organize my resumes", or "version of my resume for this job".
+description: Use when user says "manage my resumes", "track resume versions", "which resume did I send", or "master resume". Tracks versions, maintains one master resume, manages per-job tailored copies.
+category: career
 ---
 
 # Resume Version Manager
@@ -43,7 +44,7 @@ A systematic approach with:
 
 ### What is a Master Resume?
 
-A comprehensive document containing:
+A complete document containing:
 - ALL your experiences (not just recent)
 - ALL bullet points you've ever written
 - Every achievement, project, skill
@@ -165,7 +166,7 @@ Resume/
 
 **Enterprise/Corporate:**
 - Emphasizes: Scale, process, collaboration
-- Tone: Professional, structured, comprehensive
+- Tone: Professional, structured, thorough
 
 **Finance:**
 - Emphasizes: Analysis, compliance, accuracy
@@ -220,7 +221,7 @@ Resume/
 - New job or promotion
 - Completed major project
 - New skills or certifications
-- Significant achievements
+- Notable achievements
 - Awards or recognition
 
 **Quarterly Review:**

diff --git a/skills/career/salary-negotiation-prep/SKILL.md b/skills/career/salary-negotiation-prep/SKILL.md
@@ -1,6 +1,7 @@
 ---
 name: salary-negotiation-prep
-description: Research market comp bands, build a negotiation strategy, and write counter-offer scripts. Use when user says "negotiate my offer", "salary negotiation", "counter offer", "what should I ask for", "is this offer fair", "they lowballed me", or "how much should I counter".
+description: Use when user says "negotiate my offer", "salary negotiation", "counter offer", "is this offer fair", or "how much should I counter". Researches comp bands, builds strategy, writes counter scripts.
+category: career
 ---
 
 # Salary Negotiation Prep

diff --git a/skills/caveman/caveman-commit/SKILL.md b/skills/caveman/caveman-commit/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: caveman-commit
-description: >-
-  Use when user says "write a commit", "commit message", or "/commit". Conventional Commits, intent-first subject, body only when why is non-obvious.
+description: 'Use when user says "write a commit", "commit message", or "/commit". Conventional Commits, intent-first subject, body only when why is non-obvious.'
+category: caveman
 ---
 
 # Caveman Commit
@@ -11,9 +11,9 @@ Write commit messages terse and exact. Conventional Commits format. No fluff. Wh
 ## Rules
 
 **Subject line:**
-- `<type>(<scope>): <imperative summary>` — `<scope>` optional
+- `<type>(<scope>): <imperative summary>`, `<scope>` optional
 - Types: `feat`, `fix`, `refactor`, `perf`, `docs`, `test`, `chore`, `build`, `ci`, `style`, `revert`
-- Imperative mood: "add", "fix", "remove" — not "added", "adds", "adding"
+- Imperative mood: "add", "fix", "remove", not "added", "adds", "adding"
 - ≤50 chars when possible, hard cap 72
 - No trailing period
 - Match project convention for capitalization after the colon
@@ -26,8 +26,8 @@ Write commit messages terse and exact. Conventional Commits format. No fluff. Wh
 - Reference issues/PRs at end: `Closes #42`, `Refs #17`
 
 **What NEVER goes in:**
-- "This commit does X", "I", "we", "now", "currently" — the diff says what
-- "As requested by..." — use Co-authored-by trailer
+- "This commit does X", "I", "we", "now", "currently", the diff says what
+- "As requested by...", use Co-authored-by trailer
 - "Generated with Claude Code" or any AI attribution
 - Emoji (unless project convention requires)
 - Restating the file name when scope already says it
@@ -57,7 +57,7 @@ Diff: breaking API change
 
 ## Auto-Clarity
 
-Always include body for: breaking changes, security fixes, data migrations, anything reverting a prior commit. Never compress these into subject-only — future debuggers need the context.
+Always include body for: breaking changes, security fixes, data migrations, anything reverting a prior commit. Never compress these into subject-only, future debuggers need the context.
 
 ## Boundaries
 

diff --git a/skills/caveman/caveman-compress/SKILL.md b/skills/caveman/caveman-compress/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: caveman-compress
-description: >-
-  Use when user says "compress this", "shorten prompt", or "/caveman-compress". Token compression preserving intent.
+description: 'Use when user says "compress this", "shorten prompt", or "/caveman-compress". Token compression preserving intent.'
+category: caveman
 ---
 
 # Caveman Compress
@@ -40,7 +40,7 @@ cd caveman-compress && python3 -m scripts <absolute_filepath>
 - Pleasantries: "sure", "certainly", "of course", "happy to", "I'd recommend"
 - Hedging: "it might be worth", "you could consider", "it would be good to"
 - Redundant phrasing: "in order to" → "to", "make sure to" → "ensure", "the reason is because" → "because"
-- Connective fluff: "however", "furthermore", "additionally", "in addition"
+- Connective fluff: "however", `furthermore`, `additionally`, "in addition"
 
 ### Preserve EXACTLY (never modify)
 - Code blocks (fenced ``` and indented)

diff --git a/skills/caveman/caveman-help/SKILL.md b/skills/caveman/caveman-help/SKILL.md
@@ -1,12 +1,12 @@
 ---
 name: caveman-help
-description: >-
-  Use when user says "caveman help" or "caveman commands". Quick reference for Caveman modes, triggers, levels.
+description: 'Use when user says "caveman help" or "caveman commands". Quick reference for Caveman modes, triggers, levels.'
+category: caveman
 ---
 
 # Caveman Help
 
-Display this reference card when invoked. One-shot — do NOT change mode, write flag files, or persist anything. Output in caveman style.
+Display this reference card when invoked. One-shot, do NOT change mode, write flag files, or persist anything. Output in caveman style.
 
 ## Modes
 

diff --git a/skills/caveman/caveman-review/SKILL.md b/skills/caveman/caveman-review/SKILL.md
@@ -1,28 +1,28 @@
 ---
 name: caveman-review
-description: >-
-  Use when user says "review this PR", "code review", or "/caveman-review". One-line PR comments: location, problem, fix.
+description: 'Use when user says "review this PR", "code review", or "/caveman-review". One-line PR comments: location, problem, fix.'
+category: caveman
 ---
 
 # Caveman Review
 Write code review comments terse and actionable. One line per finding. Location, problem, fix. No throat-clearing.
 
 ## Rules
 
-**Format:** `L<line>: <problem>. <fix>.` — or `<file>:L<line>: ...` when reviewing multi-file diffs.
+**Format:** `L<line>: <problem>. <fix>.`, or `<file>:L<line>: ...` when reviewing multi-file diffs.
 
 **Severity prefix (optional, when mixed):**
-- `🔴 bug:` — broken behavior, will cause incident
-- `🟡 risk:` — works but fragile (race, missing null check, swallowed error)
-- `🔵 nit:` — style, naming, micro-optim. Author can ignore
-- `❓ q:` — genuine question, not a suggestion
+- `🔴 bug:`, broken behavior, will cause incident
+- `🟡 risk:`, works but fragile (race, missing null check, swallowed error)
+- `🔵 nit:`, style, naming, micro-optim. Author can ignore
+- `❓ q:`, genuine question, not a suggestion
 
 **Drop:**
 - "I noticed that...", "It seems like...", "You might want to consider..."
-- "This is just a suggestion but..." — use `nit:` instead
-- "Great work!", "Looks good overall but..." — say it once at the top, not per comment
-- Restating what the line does — the reviewer can read the diff
-- Hedging ("perhaps", "maybe", "I think") — if unsure use `q:`
+- "This is just a suggestion but...", use `nit:` instead
+- "Great work!", "Looks good overall but...", say it once at the top, not per comment
+- Restating what the line does, the reviewer can read the diff
+- Hedging ("perhaps", "maybe", "I think"), if unsure use `q:`
 
 **Keep:**
 - Exact line numbers
@@ -50,4 +50,4 @@ Drop terse mode for: security findings (CVE-class bugs need full explanation + r
 
 ## Boundaries
 
-Reviews only — does not write the code fix, does not approve/request-changes, does not run linters. Output the comment(s) ready to paste into the PR. "stop caveman-review" or "normal mode": revert to verbose review style.
+Reviews only, does not write the code fix, does not approve/request-changes, does not run linters. Output the comment(s) ready to paste into the PR. "stop caveman-review" or "normal mode": revert to verbose review style.