A collection of Agent Skills (Markdown with YAML frontmatter) for use in Cursor and similar tools. Each subdirectory is one skill; the entry point is SKILL.md.
These files are informal reference material, not reviewed software. They may be wrong, outdated, or unsafe for your environment. Use at your own risk. Verify commands, credentials, and infrastructure impact before you run anything. Nothing here is professional, legal, or security advice.
To the extent permitted by law, this repository is dedicated to the public domain under CC0 1.0 (or equivalently: no rights reserved). You may copy, modify, and distribute the content without attribution. Third-party code or vendored dependencies keep their own licenses.
skill-authoring — Follow Anthropic-style (and similar) published guidance: token efficiency and progressive disclosure. Use it to refactor skills sensibly and to fold new learning or related skills into a tighter, organized set. Most skills here were written or refactored with this playbook.
notebook — Strong fit for web research: capture sources into a structured knowledge base, then query it for narrow facts or cross-cutting synthesis. Pair it with skill-authoring when you are turning that research into a large, well-grounded skill (sources first, then distill into layered instructions).
plan — A Cursor-oriented rework of Getting Shit Done, aimed at continuability: start a fresh agent session when context gets heavy, while restartable plans in tmp/plans/ keep dependency order and verification attached to every task so the next session knows what “done” means.
| Slug | Description |
|---|---|
ansible |
Ansible playbooks and patterns for provisioning, configuration management, deployment, and multi-host orchestration. |
conversation-notes |
Workflow for saving a short summary of what was learned in a chat when the user asks for a note or recap. |
docker |
Dockerfile and Docker Compose guidance: images, multi-stage builds, Compose layout, and containerization patterns (not raw CLI—see docker-cli). |
docker-cli |
Docker CLI reference for building images, running containers, networks, volumes, and scripting orchestration. |
docx |
Creating and editing Word .docx files with an emphasis on structure, tracked changes, comments, and text extraction. |
file-change-tracker |
Convention for listing modified files as markdown links, grouped by folder, at the end of agent responses. |
foundation-models |
Apple Foundation Models (on-device LLM): Swift integration, guided generation, tools, tagging, adapters, prompts, and safety. |
github-cli |
Comprehensive gh usage for repos, issues, PRs, Actions, releases, orgs, and scripting against GitHub from the terminal. |
hummingbird |
Swift Hummingbird 2: routing, app lifecycle, middleware, streaming, SPM, and related ecosystem packages. |
ical |
Valid iCalendar (.ics) authoring per RFC 5545 (plus common extensions): folding, time zones, recurrence, and client quirks. |
ios-speech |
On-device speech on iOS: Speech framework (STT) and AVSpeech (TTS) for voice input, prompts, and voice-driven UX. |
jira-acli |
Jira work through Atlassian CLI (acli) and markdown→ADF for descriptions and comments without an MCP server. |
mac-container-virtualization |
Apple’s container tooling on macOS: CLI usage, images, networking (including Tailscale), and migration notes from Docker. |
markdown-exporter |
Converting Markdown to many formats (DOCX, PPTX, PDF, HTML, etc.) and extracting fenced code blocks. |
markdown-to-confluence |
Uploading or updating Confluence pages from Markdown via the REST API (with local sanitization/conversion helpers). |
markdown-to-pdf |
Styled PDF export from Markdown using pandoc, WeasyPrint, and CSS. |
mcp-builder |
Patterns and checklists for building MCP servers (e.g. FastMCP in Python, MCP SDK in TypeScript) with solid tool design. |
mermaid-diagrams |
Mermaid syntax for flowcharts, sequence diagrams, ER diagrams, C4, state machines, Gantt charts, and similar diagrams. |
notebook |
Building and querying document knowledge bases: acquire sources, metadata, citations, and research-style workflows. |
pi-ssh-access |
SSH to Raspberry Pi: keys, scp, tunnels, and notes on Tailscale Serve/HTTPS-style access. |
plan |
Cursor-adapted Getting Shit Done: dependency-ordered tasks, checkboxes, verification criteria, and resuming across sessions (tmp/plans/). |
pptx |
Creating and editing PowerPoint .pptx files: slides, layouts, notes, and light analysis. |
puppeteer |
Headless Chrome automation for JS-heavy pages, clean content extraction, and acquisition pipelines (pairs with notebook). |
skill-authoring |
Best-practice skill writing (Anthropic et al.): progressive disclosure and token economy; refactor skills and merge related material into one coherent skill. |
skill-catalog |
Generating catalog-style listings of skills with names and short descriptions. |
subagents |
Designing Cursor subagents: when to delegate, performance, and how subagents relate to skills, rules, and commands. |
tips |
U.S. Treasury Fiscal Data for TIPS/CPI: API usage and math for inflation-adjusted coupons, principal, and ladders. |
vision-ocr |
Text extraction from scans and images on macOS using Apple Vision (Live Text), outputting per-page text without cloud APIs. |
workspace-context |
Conventions for keeping plans, scratch notes, and exports under workspace tmp/ (or similar) subfolders. |
xcode-build |
Building and running Apple platforms with xcodebuild, xcrun simctl, devices, and UI test automation from the CLI. |
xlsx |
Spreadsheet work: .xlsx/.xlsm/CSV/TSV creation, formulas, formatting, analysis, and recalculation. |
SKILL.md— YAML frontmatter (name,description) plus pointers and core instructions.docs/,scripts/,reference/— Extra material loaded on demand whenSKILL.mdlinks to it (progressive disclosure).
Install paths depend on your tool; Cursor typically reads skills from a configured skills directory. See your editor’s documentation.