Public skills for Claude Code and Codex — privacy-scanned and generalized for anyone to install.
Skills are instruction files that extend an AI coding agent with domain-specific workflows. Install a skill and the agent knows how to handle that task type automatically.
Every skill here ships in dual-runtime format: SKILL.md for Claude Code, agents/openai.yaml companion for Codex.
Claude Code:
# copy the skill folder into your skills directory
cp -r skills/<skill-name> ~/.claude/skills/Codex:
cp -r skills/<skill-name> $CODEX_HOME/skills/Windows (PowerShell):
Copy-Item -Recurse skills\<skill-name> "$env:USERPROFILE\.claude\skills\" # Claude Code
Copy-Item -Recurse skills\<skill-name> "$env:USERPROFILE\.codex\skills\" # Codex| Skill | Description |
|---|---|
| angel-investing | Investor-side early-stage evaluation OS — 5-minute screen, deal-specific Critical Success Factors with superpower/critical-flaw scoring, three-tier diligence, valuation & SAFE/convertible/term-sheet review, cap-table & dilution modeling, portfolio fit, and investment-memo generation |
| consulting-engagements | Run a solo/small-practice B2B consulting engagement end to end — discovery, proposals/SOW, source-ledger research, issue trees, executive briefs/decks/decision memos, QA gates, and confidentiality controls, with a fabrication firewall and human-approval gates |
| developing-nonfiction-books | Develop substantive nonfiction books from concept to publication-ready manuscript — book diagnosis, architecture selection, claim-level evidence control with A–D confidence levels, seven staged revision modes, chapter-by-chapter collaboration protocol, quality gates, and calibrated readiness labels |
| editorial-writer | Publication-grade long-form articles and multi-part series — topic-vs-idea triage, claim-and-evidence ledger before prose, claim-type taxonomy with a corroboration rule, argument architecture and outline gate, post-draft evidence audit, reverse-outline pass, hard gates separated from optimization metrics, headline-integrity scoring, publish packet and corrections log, plus an Arabic (Saudi/GCC) editorial layer with a geographic-scope gate |
| gemini-notebook-enterprise | Automate Google Gemini Notebook Enterprise (NotebookLM Enterprise) via the documented Discovery Engine Notebook API — notebook/source inventory, ingestion-failure diagnosis, OAuth+IAM model, agent/MCP integration tiers, and a capability contract that refuses to guess an undocumented Q&A endpoint |
| latex-professional | Create, edit, compile, and package professional LaTeX documents — compile-error triage, BibTeX/BibLaTeX/Biber bibliography repair, math-meaning preservation, tables/figures/equations, and arXiv/IEEE/ACM/Springer/Elsevier/thesis submission packaging |
| meta-prompt-engineering | Write and audit prompts in two modes — one copy-ready prompt for a target model with per-model adaptation, or a persistent agent instruction layer as a testable contract: orchestration spine, explicit trust ladder, one-strongest-enforcement-point placement, fifteen structural failure modes, eight meta-prompt templates (orchestrator, executor, coordinator, delegation, injection-resistant analyst, recovery, judge, optimizer), eval pyramid with trajectory metrics and holdout release gates, and a stdlib CI linter for budget/duplication/capability-sync/test-ID coverage |
| obsidian-writing | Write, edit, and organize an Obsidian vault — wikilinks/callouts/frontmatter, Dataview/Tasks/Bases dashboards, Canvas files, Templater automation, citation workflows, REST API/MCP control, and an optional anti-drift vault system (one-note-one-home routing, self-routing templates, project boards) |
| powershell-safe-vars | PowerShell footgun catalog for AI-agent tool calls — automatic-variable collisions, quoting/heredoc traps, null-handling bugs, and the safe pattern for each |
| repo-stewardship | Safely acquire, inspect, sync, validate, and prepare Git repositories for review or delivery — permission tiers, quality gates, changelog/SemVer discipline, PR readiness reports |
| research-grounding | Forces a current web search before any recommendation, comparison, price, version, or "state-of-the-art" claim is presented as solid — search-and-cite workflow, red flags, rationalizations table |
| rust-engineering | Production-grade Rust engineering guidelines — ownership, type-driven design, error handling, async/concurrency, testing, security, and performance |
| startup-consulting | Evidence-based startup advisory from idea validation to scaling — stage classification, problem/customer discovery, bottom-up market sizing, unit economics, GTM, pricing, pitch-deck review, fundraising prep, and the scaling gate |
| tauri-engineering | Tauri v2 engineering — guest-host architecture, IPC commands, ACL security, plugins, binary optimization, auto-updates, and CI/CD |
| test-backend | Stack-agnostic back-end/API testing — Testing Trophy unit+integration, contract testing (Pact/Schemathesis), data invariants, resilience/chaos, OWASP API Top 10, load/SLO |
| test-ui | Stack-agnostic UI testing — visual regression, responsive + cross-browser, WCAG 2.2 AA accessibility, component/state matrix, i18n/locale-integrity, design-token fidelity |
| test-ux | Stack-agnostic UX testing — Nielsen heuristic evaluation, task-based cognitive walkthroughs, journey/state analysis, content/i18n comprehension, real-user test-plan generator |
| ux-engineering | UX engineering and product design — JTBD research, information architecture, WCAG 2.2 AA accessibility, design systems, interaction design, HEART/SUS metrics |
Each skill passes a privacy scan and generalization pass before publishing. Process details: docs/PUBLISHING.md.
Adapt freely — replace any <placeholder> values with your own config. PRs welcome if they pass the quality bar.