/agent skills

Official Angular skill from angular/skills — generates Angular code and architectural guidance for components, signals (signal, linkedSignal, resource, httpResource), forms (signal / template / reactive), routing, DI, SSR, ARIA, animations, styling (component styles, Tailwind CSS), testing, naming conventions, and CLI tooling. Reads project Angular version first; loads the matching reference per t

Anthropic's official mcp-builder skill from anthropics/skills — a four-phase workflow (research & planning → implementation → review & test → evaluations) for building MCP servers in TypeScript (recommended) or Python (FastMCP). Loads reference files for best practices, language guides, and an evaluation generator.

Anthropic's official PPTX skill from anthropics/skills — create, edit, read, split, and combine .pptx and .potx files via pptxgenjs (create), XML unzip-rezip (edit), markitdown (read), and LibreOffice (validate, convert). Fixed stack: pptxgenjs + markitdown[pptx] + Pillow + defusedxml + lxml + LibreOffice (soffice) + pdftoppm.

Operator-authored Agent Skill: audit a URL, screenshot, Figma frame, or component against Adam Wathan and Steve Schoger's Refactoring UI, then return a ranked fix report. Not an Anthropic demo skill.

Anthropic's official frontend-design skill from anthropics/skills — guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Two-pass plan-then-build grounded in the brief's subject matter. Names five AI-generated default looks the agent should refuse as defaults.

No standalone bash skill in anthropics/skills. Bash execution for Claude lives in two places: the native Bash tool in Claude Code, and the computer use toolset (computer_toolset_20260801) in anthropic-quickstarts/computer-use-demo.

Anthropic's official DOCX skill from anthropics/skills — create, read, edit, and manipulate Word documents (.docx / .dotx) via docx-js for creation and a Python unzip/edit/rezip pipeline for edits; comments and tracked changes via bundled helpers.

Anthropic's official skill-creator from anthropics/skills — a meta-procedure for creating new skills, improving existing skills, running evals to test them, benchmarking with variance analysis, and optimizing SKILL.md frontmatter description for triggering accuracy. Iterative loop: capture intent → interview → write SKILL.md → test prompts → grade → reviewer → improve → repeat.

Anthropic's official XLSX skill from anthropics/skills — create, read, edit, and analyze spreadsheets (.xlsx / .xlsm / .xltx / .csv / .tsv) via openpyxl, pandas, and markitdown. Includes a mandatory recalc.py step that runs LibreOffice headless to bake formulas before delivery.

Google-published Agent Skill for the Gemini API on Agent Platform (formerly Vertex AI). Drives the Gen AI SDK across Python, JS/TS, Go, Java, and C# — picks the right model, picks ADC vs API-key auth, refuses legacy SDKs. Not a generic 'use Gemini' prompt.

Meta-skill from NVIDIA/skills — routes NVIDIA-related requests to the right NVIDIA-published skill across 300+ entries. Triggers on NVIDIA products, CUDA, cuDF, Jetson, NeMo, Omniverse, TAO, VSS, DOCA, and so on, then checks the live NVIDIA catalog via `npx skills add nvidia/skills --list` (or fallback URLs) before recommending an install. Never installs without confirmation.

obra/superpowers methodology skill. Forces a HARD-GATE human-approval before any implementation. Classifies the request as spike / bounded / architectural, asks questions one at a time, writes a spec for the architectural path. Not an Anthropic demo skill.

Superpowers workflow for implementing a written plan step-by-step in a single session, with verification at each step and a clear stop-on-blocker policy. Use when you have a written implementation plan and want the agent to walk it to completion.

obra/superpowers methodology skill. Use when hitting any bug, test failure, or unexpected behavior, before proposing fixes. Four-phase root-cause process: investigate → pattern → hypothesis → implement. No fixes without root cause first. Not an Anthropic demo skill.

obra/superpowers methodology skill. The iron law is one line: NO PRODUCTION CODE WITHOUT A FAILING TEST FIRST. Enforces red-green-refactor on every feature, bugfix, and refactor, with explicit red-flag and rationalization tables to talk the agent out of skipping the watch-it-fail step. Not an Anthropic demo skill.

obra/superpowers methodology skill. Turns an approved spec or requirements doc into a bite-sized implementation plan a junior engineer with no project context can follow, task by task. Not an Anthropic demo skill.

obra/superpowers meta-skill. Creates and edits SKILL.md files using TDD applied to documentation — RED (baseline behavior with subagents), GREEN (write the skill), REFACTOR (close rationalization loopholes). Not an Anthropic demo skill.

Redis's redis-semantic-cache skill from redis/agent-skills — cache-aside pattern for LLM responses on Redis Cloud LangCache. Calls search/set via SDK or REST, tunes the similarity threshold (0.95+ strict / 0.9 default / 0.8 loose), splits one app's LLM workloads into separate caches per task type, and filters with custom attributes. Use when wrapping an LLM call (OpenAI, Anthropic) to cut cost and

Resend's email-best-practices skill from resend/resend-skills — guidance for deliverable, compliant, user-friendly email. Covers SPF/DKIM/DMARC authentication, transactional vs marketing, CAN-SPAM/GDPR/CASL compliance, double opt-in, idempotent send + retry, webhook event handling, suppression lists, accessibility (alt text, headings, contrast, screen readers). Use when emails hit spam, bounce rat

Procedural rules for React component architecture — kill boolean prop proliferation, prefer compound components, lift state into providers, and compose internals. Notes React 19 API changes (no more forwardRef).

Discovery skill from vercel-labs/skills — when the user asks 'how do I do X' or 'find a skill for X', check the skills.sh leaderboard, then run `npx skills find`, verify install counts and source reputation, and present install commands. Does not install without confirmation.

Procedural rules for React and Next.js performance from vercel-labs/agent-skills — kill request waterfalls, shrink bundles, parallelize server fetches, stop unnecessary re-renders. Prioritised by impact, 70 rules across 8 categories.

Review UI code against Vercel's Web Interface Guidelines — fetches the live rule set from a raw URL on every run, audits accessibility, focus states, forms, animation, typography, images, performance, navigation, dark mode, touch, and i18n. Outputs terse file:line findings.