{"id":47790739,"url":"https://github.com/psenger/ai-agent-skills","last_synced_at":"2026-04-25T08:04:00.473Z","repository":{"id":344307695,"uuid":"1181265319","full_name":"psenger/ai-agent-skills","owner":"psenger","description":"Production-ready AI agent skills for Claude Code, Codex, Cursor, and the Agent Skills Open Standard","archived":false,"fork":false,"pushed_at":"2026-03-22T07:27:56.000Z","size":232,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-03-22T18:52:28.743Z","etag":null,"topics":["agent-skills","ai-agent","api-design","api-security","architectural-review","claude","claude-code","codex","cursor","developer-tools","git","llm","obsidian","openapi","prompt-engineering","rest-api"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/psenger.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-03-13T23:38:39.000Z","updated_at":"2026-03-22T07:29:36.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/psenger/ai-agent-skills","commit_stats":null,"previous_names":["psenger/ai-agent-skills"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/psenger/ai-agent-skills","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/psenger%2Fai-agent-skills","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/psenger%2Fai-agent-skills/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/psenger%2Fai-agent-skills/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/psenger%2Fai-agent-skills/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/psenger","download_url":"https://codeload.github.com/psenger/ai-agent-skills/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/psenger%2Fai-agent-skills/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31360807,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-03T15:19:21.178Z","status":"ssl_error","status_checked_at":"2026-04-03T15:19:20.670Z","response_time":107,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-skills","ai-agent","api-design","api-security","architectural-review","claude","claude-code","codex","cursor","developer-tools","git","llm","obsidian","openapi","prompt-engineering","rest-api"],"created_at":"2026-04-03T15:37:08.256Z","updated_at":"2026-04-25T08:04:00.463Z","avatar_url":"https://github.com/psenger.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cp align=\"center\"\u003e\n  \u003ch1 align=\"center\"\u003eai-agent-skills\u003c/h1\u003e\n  \u003cp align=\"center\"\u003e\n    A curated collection of production-ready AI agent skills\u003cbr/\u003e\n    for Claude Code, Codex, Cursor, and any tool that supports the\u003cbr/\u003e\n    \u003ca href=\"https://agentskills.io/specification\"\u003eAgent Skills Open Standard\u003c/a\u003e.\n  \u003c/p\u003e\n  \u003cp align=\"center\"\u003e\n    \u003ca href=\"https://github.com/psenger/ai-agent-skills/blob/main/LICENSE\"\u003e\u003cimg src=\"https://img.shields.io/github/license/psenger/ai-agent-skills?style=flat-square\" alt=\"License\"\u003e\u003c/a\u003e\n    \u003ca href=\"https://github.com/psenger/ai-agent-skills/issues\"\u003e\u003cimg src=\"https://img.shields.io/github/issues/psenger/ai-agent-skills?style=flat-square\" alt=\"Issues\"\u003e\u003c/a\u003e\n    \u003ca href=\"https://github.com/psenger/ai-agent-skills/stargazers\"\u003e\u003cimg src=\"https://img.shields.io/github/stars/psenger/ai-agent-skills?style=flat-square\" alt=\"Stars\"\u003e\u003c/a\u003e\n  \u003c/p\u003e\n\u003c/p\u003e\n\n---\n\n## What Is This?\n\n**ai-agent-skills** is a library of reusable skills that teach AI coding agents _how_ to do specialised tasks — consistently, every time. Each skill is a self-contained folder with structured instructions, reference material, and examples that an agent loads on demand.\n\nThink of it as a playbook: you define the process once, and the agent follows it whenever the task comes up.\n\n---\n\n## Skills\n\n| Skill | Type | Description |\n|---|---|---|\n| **[vault-scribe](skills/vault-scribe/)** | `/vault-scribe` | Converts transcripts, meeting notes, brainstorming sessions, strategy docs, and rough notes into polished Obsidian vault Markdown — GitHub-compatible by default, with type-aware frontmatter schemas |\n| **[agentic-skeleton-dir-structure](skills/agentic-skeleton-dir-structure/)** | `/agentic-skeleton-dir-structure` | Scaffolds production-ready directory structures for agentic AI projects using Agent-OS v3 (Builder Methods) — supports single repos, mono-repos, multi-language repos, any platform, any language |\n| **[git-commit-pr-message](skills/git-commit-pr-message/)** | `/git-commit-pr-message` | Generates Conventional Commits messages, PR titles/descriptions, and Keep a Changelog v1.1.0 entries — with sensitive content scanning, GitHub/Jira ticket linking, and release workflow |\n| **[design-critique](skills/design-critique/)** | `/design-critique` | Structured design critique and plan stress-testing — acts as a relentless interviewer using pre-mortem, red teaming, and ATAM techniques to challenge technical architectures, product plans, and feature designs exhaustively |\n| **[arch-lens](skills/arch-lens/)** | `/arch-lens` | Seven-step interactive architectural review anchored in Ousterhout's deep-module principle — explores a codebase for shallow modules, hidden coupling, and testability seams, spawns parallel sub-agents to design competing interfaces, then writes a structured RFC action file readable by GitHub MCP or ROVO (Jira) MCP |\n| **[review-api-design](skills/review-api-design/)** | `/review-api-design` | Reviews REST API designs during the planning phase against security, resilience, design, and operational best practices — produces structured findings with severity levels, source citations, and a readiness assessment |\n| **[create-a-skill](skills/create-a-skill/)** | `/create-a-skill` | Create new agent skills from scratch, modify and improve existing skills, and measure skill performance — interviews the user, drafts SKILL.md with bundled resources, runs evals, benchmarks, iterates on feedback, optimises description triggering, and packages distributable `.skill` files |\n| **[handoff](skills/handoff/)** | `/handoff` | Saves or loads a structured JSON snapshot of session state so work can resume cleanly in a new session or be delegated to a sub-agent — better than `/compact` because the schema forces every field to be explicit |\n\n### vault-scribe\n\nYour Obsidian vault assistant. Turns unstructured input into well-organised, searchable notes with proper YAML frontmatter, callout blocks, and consistent formatting.\n\n**Supports five note types:**\n\n| Note Type | Use Case |\n|---|---|\n| `article` | Knowledge base articles, guides, reference docs |\n| `meeting` | Meeting notes, 1:1s, standups, retrospectives |\n| `brainstorming` | Ideation sessions — solo with AI, one-on-one, or group |\n| `strategy` | Versioned plans, OKRs, and living strategy documents |\n| `deep-research` | In-depth investigations with multiple sources |\n\n**Key features:**\n- GitHub-Flavored Markdown first, Obsidian extensions when needed\n- Type-aware YAML frontmatter with enforced schemas\n- Automatic transcript appendix formatting\n- GFM Alerts for cross-platform callout blocks\n- Quality checklist validation before output\n- Reference files for frontmatter schemas, callout types, embed syntax, and Markdown compatibility\n\n```\nvault-scribe/\n├── SKILL.md                          Workflow + quality checklist\n├── references/\n│   ├── FRONT-MATTER.md               Frontmatter schemas for all note types\n│   ├── CALLOUTS.md                   GFM Alerts + Obsidian callout reference\n│   ├── EMBEDS.md                     Image + embed syntax (GFM + Obsidian)\n│   └── MARKDOWN-SYNTAX.md           Links, tags, math, diagrams, footnotes\n└── examples/\n    ├── article-example.md            Article with callouts + code blocks\n    ├── meeting-brainstorm-example.md  Brainstorming session\n    └── transcript-example.md         Article with transcript appendix\n```\n\n### agentic-skeleton-dir-structure\n\nYour project scaffolding assistant. Interactively builds production-ready directory structures for agentic AI projects, with [Agent-OS v3](https://github.com/buildermethods/agent-os) by Builder Methods and [Spec-Driven Development (SDD)](https://buildermethods.com/library/spec-driven-development-claude-code) baked in.\n\n**How it works:**\n\n1. **Detects context** — checks if the current directory already has files and warns you before overwriting anything\n2. **Asks 6 questions** (one at a time) — repo pattern, platform type, languages, IaC tool, target platform, agent tooling\n3. **Shows a summary table** and waits for your confirmation before creating anything\n4. **Scaffolds everything** — directories, `CLAUDE.md`, `.claude/` config, `agent-os/` structure, IaC layout, seed files\n5. **Guides next steps** — Agent-OS installation, `/plan-product`, and the SDD workflow loop\n\n**Supports any combination of:**\n\n| Dimension | Options |\n|---|---|\n| Repo pattern | Single Repo, Mono-Repo, Multi-Language Mono-Repo |\n| Platform | Frontend, Backend, Full-Stack, Middleware, Agents/AI |\n| Language | TypeScript, Python, Go, Java, Rust, Ruby, C#, and more |\n| IaC | Terraform, Pulumi, CDK, Bicep, CloudFormation, Helm, Ansible |\n\n**What gets created:**\n\n- `CLAUDE.md` — project instructions loaded every Claude Code session\n- `.claude/` — agents, skills, commands, hooks directories\n- `agent-os/` — standards, specs, product context (mission, roadmap, tech stack)\n- `src/` or `apps/` + `packages/` — language-specific source layout\n- `iac/` — IaC structure matching your chosen tool\n- `deploy/` — CI/CD pipelines, Docker, deploy scripts\n- `docs/` — architecture decision records, API docs, runbooks\n\n```\nagentic-skeleton-dir-structure/\n├── SKILL.md                          Interactive workflow + quality checklist\n├── references/\n│   ├── repo-patterns.md              Language + platform source layouts (7 languages)\n│   ├── iac-patterns.md               IaC by tool (7 tools) + CI/CD + env promotion\n│   └── agent-os-guide.md             Agent-OS install, SDD methodology, commands\n└── examples/\n    ├── single-repo-typescript.md     Completed single repo TypeScript API scaffold\n    └── mono-repo-fullstack.md        Completed mono-repo full-stack scaffold\n```\n\n### git-commit-pr-message\n\nYour commit and PR workflow assistant. Generates professional git commit messages, pull request titles and descriptions, changelog entries, and handles releases — all following industry-standard conventions.\n\n**What it does:**\n\n1. **Scans for sensitive content** (API keys, tokens, passwords, private keys) — mandatory gate before any commit\n2. **Asks for ticket references** — supports GitHub Issues (all 9 closing keywords) and Jira (pattern-matched ticket keys)\n3. **Generates commit messages** — Conventional Commits format with type, scope, subject, body, and footer\n4. **Updates CHANGELOG.md** — Keep a Changelog v1.1.0 with all six section types (Added, Changed, Deprecated, Removed, Fixed, Security)\n5. **Creates pull requests** — via `gh` CLI or GitHub MCP, with summary, ticket links, changes, and test plan\n6. **Cuts releases** — renames Unreleased to versioned section, adds comparison links, optionally creates git tags\n\n**Key features:**\n- Conventional Commits with type, scope, and imperative mood enforcement\n- All 9 GitHub closing keywords (`close/closes/closed`, `fix/fixes/fixed`, `resolve/resolves/resolved`)\n- Jira ticket key detection by pattern (`PROJ-1234`) — no prefix needed\n- Keep a Changelog v1.1.0 with comparison links at bottom of file\n- Sensitive content scanning with line-level reporting\n- User confirmation gates — never commits, pushes, or creates PRs without asking\n- Skills v2.0 compliant with `disable-model-invocation`, `allowed-tools`, `argument-hint`\n\n```\ngit-commit-pr-message/\n├── SKILL.md                          Workflow (9 steps) + behavioural rules\n└── references/\n    └── examples.md                   Commit, PR, changelog, ticket, and scan examples\n```\n\n### design-critique\n\nYour design review sparring partner. Stress-tests technical architectures, product plans, and feature designs using structured interviewing techniques drawn from pre-mortem analysis, red teaming, and ATAM (Architecture Tradeoff Analysis Method).\n\n**What it does:**\n\n1. **Orients silently** — explores the codebase or relevant files before asking anything\n2. **Anchors the session** — establishes scope with a single opening question\n3. **Drills relentlessly** — one question at a time, following the highest-risk thread first\n4. **Surfaces hidden assumptions** — names what's unstated and forces trade-off articulation\n5. **Closes with a summary** — what held up, what didn't, and what needs resolution before proceeding\n\n**Question patterns it uses:**\n\n| Pattern | Purpose |\n|---|---|\n| What happens when X fails? | Failure modes |\n| What does the alternative look like? | Trade-off articulation |\n| How would you know if this is wrong? | Falsifiability |\n| What's the cost of reversing this? | Reversibility |\n| Walk me through the worst case | Pre-mortem |\n| What quality attribute does this sacrifice? | ATAM tradeoff probe |\n\n**Skills 2.0:** `allowed-tools: Read Grep Glob` — `argument-hint: [topic, file, or artifact to critique]` — auto-invokes on trigger phrases; no external tools required\n\n```\ndesign-critique/\n└── SKILL.md                          Interviewing principles, question patterns, session flow\n```\n\n### arch-lens\n\nYour architectural review assistant. Analyses a codebase through the lens of Ousterhout's deep-module principle — a deep module has a small interface hiding a large implementation, making it testable at the boundary and navigable by AI without reading internals.\n\nDetection is organic, not mechanical: an Explore sub-agent navigates the codebase the way a developer would. The confusion it encounters, the files it has to bounce between, the test boundaries it can't find — that friction IS the signal. No checklists.\n\n**What it does:**\n\n1. **Explores organically** — spawns an Explore sub-agent that navigates the codebase naturally, recording friction: concept scatter, shallow interfaces, unreachable test seams, hidden orchestration, and integration risk at module boundaries\n2. **Presents candidate clusters** — groups friction observations into named clusters, each with: modules involved, coupling reason, co-owners, call patterns, shared types, dependency category, and existing tests that a boundary test would replace\n3. **Asks what to explore** — single open question; user picks a cluster and directs the angle\n4. **Frames the problem space** — principle violated, current interface, dependency category confirmed, blast radius, and what tests currently have to reach through to exercise the behaviour\n5. **Spawns parallel sub-agents** — 3–4 agents in parallel, each given an independent technical brief and a distinct design constraint (minimise, maximise flexibility, optimise for common caller, ports \u0026 adapters)\n6. **Presents designs and recommends** — interface signature, usage example, hidden complexity, dependency strategy, and trade-offs per design; compared in a table and prose; followed by a strong opinionated recommendation or named hybrid\n7. **Writes an RFC action file** — `arch-rfcs-YYYY-MM-DD.md` at the project root; one RFC per finding with Problem, Proposed Interface, Dependency Strategy, Testing Strategy, and Implementation Recommendations — structured for direct consumption by GitHub MCP or ROVO (Jira) MCP\n\n**Dependency categories** — every candidate cluster is classified into one of four categories that determine the testing strategy:\n\n| Category | What it means | Testing approach |\n|---|---|---|\n| In-process | Pure computation, no I/O | Test directly — no adapters |\n| Local-substitutable | Infrastructure with a high-fidelity stand-in | Test with PGLite, in-memory FS, etc. |\n| Remote but owned | Your services across a network boundary | Ports \u0026 adapters — in-memory adapter for tests |\n| True external | Third-party services you don't control | Mock at the boundary |\n\n**Testing strategy** — replace, don't layer. Old unit tests on shallow modules become waste once boundary tests exist — delete them. New tests assert on observable outcomes through the public interface, not internal state.\n\n**Sub-agent design constraints:**\n\n| Agent | Constraint |\n|---|---|\n| Agent 1 | Minimise — 1–3 entry points max, every param essential |\n| Agent 2 | Maximise flexibility — support extension without caller changes |\n| Agent 3 | Optimise for the common caller — make the default case trivial |\n| Agent 4 | Ports \u0026 adapters — pure domain interface, all infrastructure injected |\n\n**Skills 2.0:** `allowed-tools: Read Grep Glob Write Bash(git *)` — `argument-hint: [path/to/scope]` — auto-invokes on trigger phrases; requires git for churn analysis\n\n```\narch-lens/\n├── SKILL.md                          Workflow summary table + behavioural rules (66 lines)\n└── references/\n    ├── WORKFLOW.md                   Full step-by-step detail and Explore agent prompt\n    ├── DETECTION-PATTERNS.md         Friction vocabulary, dependency categories, testing strategy\n    ├── INTERFACE-DESIGN.md           Sub-agent brief template, design constraints, comparison format\n    └── RFC-FILE-FORMAT.md            Action file format, effort/priority/label mapping, full example\n```\n\n### review-api-design\n\nYour API design review assistant. Vets REST API designs during the planning phase — before a single line of code is written. Produces structured review documents with severity-rated findings, source citations, and a readiness assessment.\n\n**What it does:**\n\n1. **Gathers context** — asks about domain, consumers, scale, auth requirements, deployment, and team experience (skips questions already answered by the input)\n2. **Loads relevant references** — selectively reads from 10 domain-specific checklists based on what the design needs\n3. **Conducts systematic review** — evaluates against security, resilience, design principles, payloads, extensibility, communication patterns, gateways, and operational best practices\n4. **Produces a structured review** — summary table, detailed findings (What/Why/Recommendation with source citations), \"What's Missing\" gap analysis, and readiness assessment\n\n**Review domains (10 reference files):**\n\n| Domain | What It Covers |\n|--------|---------------|\n| Design Principles | Naming, versioning, CRUD, idempotency, health checks, tracing, parameters, ID exposure |\n| Payloads \u0026 Errors | Response structure, pagination, RFC 9457 errors, identifiers, content negotiation |\n| Security (Auth) | Zero trust, OAuth 2.0/2.1, RBAC/ABAC, MFA/passkeys, JWT, rate limiting, sessions, risk-based security |\n| Security (Defense) | Enumeration, information disclosure, input validation, CORS, CSRF, security headers, OWASP API Top 10 |\n| Extensibility | Fixed vs variable arity, metadata escape hatches, SOLID principles, response evolution, Hyrum's Law |\n| Resilience | Retries, circuit breakers, timeouts, bulkheads, caching, observability, SLIs/SLOs |\n| Communication Patterns | REST vs GraphQL vs WebSockets vs SSE — when to use each, hybrid architectures |\n| API Gateways | Gateway patterns, product comparison, when to use/skip |\n| Human Aspect | Adoption, documentation, NFRs, testing strategy |\n| Pragmatism | Dependencies, framework lock-in, build vs buy |\n\n**Invocation note:** This skill works best when invoked explicitly via `/review-api-design`. It may also activate during plan mode when API design decisions are being made, but explicit invocation is more reliable.\n\n```\nreview-api-design/\n├── SKILL.md                          Workflow + output format + example\n├── evals/\n│   └── evals.json                    3 test cases\n└── references/\n    ├── design-principles.md          Naming, versioning, CRUD, parameters\n    ├── design-extensibility.md       Arity, metadata, SOLID, response evolution\n    ├── payloads-errors.md            Response structure, pagination, errors, IDs\n    ├── security-auth.md              Identity, auth, tokens, trust boundaries\n    ├── security-defense.md           Enumeration, CSRF, CORS, info disclosure\n    ├── resilience.md                 Retries, circuit breakers, observability\n    ├── api-communication-patterns.md REST vs GraphQL vs WebSockets vs SSE\n    ├── api-gateways.md               Gateway patterns and product comparison\n    ├── human-aspect.md               Adoption, documentation, NFRs\n    ├── pragmatism.md                 Dependencies, lock-in, build vs buy\n    └── sources.md                    Consolidated references (cited in findings)\n```\n\n### create-a-skill\n\nYour skill authoring assistant. Walks you through the full lifecycle of creating, testing, and shipping an agent skill — from initial interview through packaging a distributable `.skill` file.\n\n**What it does:**\n\n1. **Gathers requirements** — interviews you about the skill's purpose, triggers, output format, edge cases, and dependencies; researches the domain via web search and MCPs\n2. **Drafts the skill** — writes SKILL.md with proper frontmatter, progressive disclosure, bundled scripts, and reference files\n3. **Tests with evals** — spawns parallel runs (with-skill vs baseline), drafts assertions, grades outputs, and aggregates benchmarks\n4. **Iterates on feedback** — launches an interactive viewer for qualitative review, reads your feedback, and rewrites the skill\n5. **Optimises description** — generates trigger eval queries, runs an automated optimisation loop with train/test split to maximise triggering accuracy\n6. **Packages** — validates and creates a `.skill` zip file ready for distribution\n\n**Key features:**\n- Detailed user interview before writing a single line\n- Web research for unfamiliar domains\n- Quantitative eval loop with grading, benchmarking, and analyst pass\n- Interactive HTML viewer for qualitative review\n- Blind A/B comparison between skill versions (advanced)\n- Description optimisation with train/test split to prevent overfitting\n- Skill 2.0 compliant output\n\n**License note:** This skill is a derivative work incorporating material from Anthropic's [skill-creator](https://github.com/anthropics/skills) (Apache 2.0) and Matt Pocock's [write-a-skill](https://github.com/mattpocock/skills) (MIT). See `skills/create-a-skill/NOTICE` and `skills/create-a-skill/LICENSE.txt` for details.\n\n```\ncreate-a-skill/\n├── SKILL.md                          Workflow (6 phases) + writing guide\n├── agents/\n│   ├── grader.md                     Assertion evaluation against outputs\n│   ├── comparator.md                 Blind A/B comparison\n│   └── analyzer.md                   Post-hoc analysis + benchmark patterns\n├── assets/\n│   └── eval_review.html              Trigger eval review UI template\n├── eval-viewer/\n│   ├── generate_review.py            Interactive result viewer server\n│   └── viewer.html                   Viewer HTML template\n├── references/\n│   └── schemas.md                    JSON schemas for all data structures\n├── scripts/\n│   ├── quick_validate.py             SKILL.md validation\n│   ├── package_skill.py              .skill file packaging\n│   ├── run_eval.py                   Trigger testing\n│   ├── run_loop.py                   Description optimisation loop\n│   ├── improve_description.py        Description improvement\n│   ├── aggregate_benchmark.py        Benchmark aggregation\n│   ├── generate_report.py            HTML report generation\n│   └── utils.py                      Shared utilities\n├── LICENSE.txt                       Apache License 2.0\n└── NOTICE                            Attribution notice\n```\n\n---\n\n## Installation\n\n### Via npx (works with Claude Code, Codex, Cursor)\n\nInstall all skills:\n\n```bash\nnpx skills add psenger/ai-agent-skills\n```\n\nList all available skills:\n\n```bash\nnpx skills list psenger/ai-agent-skills\n```\n\nInstall a specific skill:\n\n```bash\nnpx skills add psenger/ai-agent-skills --skill vault-scribe\n```\n\nInstall locally for customisation:\n\n```bash\nnpx skills add psenger/ai-agent-skills --skill vault-scribe --local\n```\n\n### Via Claude Code Marketplace\n\n```bash\nclaude plugin marketplace add psenger/ai-agent-skills\nclaude plugin install vault-scribe@psenger-skills-marketplace\n```\n\n### Manual Installation\n\nClone the repo and copy the skill folder:\n\n```bash\n# Global (available in all projects)\ncp -r skills/vault-scribe ~/.claude/skills/vault-scribe\n\n# Local (project-specific)\ncp -r skills/vault-scribe .claude/skills/vault-scribe\n```\n\n\u003e **Global vs Local**\n\u003e\n\u003e **Global** (`~/.claude/skills/`) — skills that apply everywhere.\n\u003e **Local** (`.claude/skills/`) — project-specific customisations. Local skills override global skills of the same name.\n\n---\n\n## Usage\n\nOnce installed, skills activate automatically based on your request. You can also invoke them directly:\n\n### vault-scribe\n\n```\n/vault-scribe article\n/vault-scribe meeting\n/vault-scribe brainstorming\n/vault-scribe strategy\n/vault-scribe deep-research\n```\n\nOr describe what you need — the skill triggers on context:\n\n```\n\"Turn this transcript into an Obsidian note\"\n\"Write up meeting notes from today's standup\"\n\"Create a strategy doc for the Q3 roadmap\"\n```\n\n### agentic-skeleton-dir-structure\n\n```\n/agentic-skeleton-dir-structure\n/agentic-skeleton-dir-structure single\n/agentic-skeleton-dir-structure mono\n/agentic-skeleton-dir-structure multi-lang\n```\n\nOr describe what you need — the skill triggers on context:\n\n```\n\"Set up a new project for an agentic AI service\"\n\"Scaffold a mono-repo for my full-stack TypeScript app\"\n\"Create a directory structure for this project\"\n\"Initialize an Agent-OS project layout\"\n```\n\nPass a repo pattern as an argument to skip the first question. Without arguments, the skill walks you through all six questions interactively.\n\n### git-commit-pr-message\n\n```\n/git-commit-pr-message commit\n/git-commit-pr-message pr\n/git-commit-pr-message changelog\n/git-commit-pr-message release\n```\n\nOr describe what you need — the skill triggers on context:\n\n```\n\"Commit these changes\"\n\"Create a PR for this branch\"\n\"Update the changelog\"\n\"Cut a release for v1.2.0\"\n```\n\nNote: This skill has `disable-model-invocation: true`, so it will only activate when you explicitly invoke it — it will never auto-trigger during normal conversation.\n\n### design-critique\n\n```\n/design-critique\n```\n\nOr trigger it naturally:\n\n```\n\"Grill me on this architecture\"\n\"Stress-test this plan\"\n\"Pre-mortem this feature design\"\n\"Red team my approach\"\n\"Critique this\"\n```\n\nThe skill self-directs toward relevant context — if it has file access, it reads the codebase silently before asking its first question.\n\n### arch-lens\n\n```\n/arch-lens\n/arch-lens src/payments\n```\n\nOr trigger it naturally:\n\n```\n\"Arch review this codebase\"\n\"Find shallow modules\"\n\"Surface coupling and testability issues\"\n\"Run an Ousterhout review on src/\"\n\"Find architectural friction\"\n\"Audit the module depth\"\n```\n\nPass an optional path to scope the analysis to a specific directory. Without arguments, the skill analyses the full repository. The skill walks you through all seven steps interactively — it will not proceed past candidate confirmation or interface selection without your input. The final output is an `arch-rfcs-YYYY-MM-DD.md` file at the project root ready to action with your GitHub or Jira MCP tooling.\n\n### review-api-design\n\n```\n/review-api-design\n/review-api-design POST /users, GET /users/{id}, DELETE /users/{id}\n```\n\nOr trigger it naturally:\n\n```\n\"Review my API\"\n\"API design review\"\n\"Vet this REST contract\"\n\"Check my endpoints\"\n\"Is this endpoint structure any good?\"\n```\n\nPass an endpoint list or OpenAPI spec as an argument, or paste it in a follow-up message. For vague verbal descriptions (\"I'm building an API for X\"), the skill asks clarifying questions before producing a review. The output is a structured review document with severity-rated findings and a readiness assessment.\n\n**Note:** This skill works most reliably when invoked explicitly with `/review-api-design`. It may auto-trigger during plan mode conversations about API design, but explicit invocation is recommended.\n\n### create-a-skill\n\n```\n/create-a-skill\n```\n\nOr trigger it naturally:\n\n```\n\"I want to make a skill for X\"\n\"Turn this into a skill\"\n\"Create a skill that does Y\"\n\"Write a skill for managing Z\"\n\"Help me build a new skill\"\n```\n\nThe skill interviews you about requirements before writing anything. It handles the full lifecycle — from initial draft through eval, iteration, description optimisation, and packaging.\n\n### handoff\n\nCaptures the complete state of your current session to a structured JSON snapshot so you can resume cleanly in a new session, switch task phases, or delegate work to a sub-agent without losing context.\n\n**Two modes:**\n\n| Mode | When to use |\n|---|---|\n| CREATE | Session approaching 300–400k tokens, switching phases, delegating to a sub-agent, or starting fresh with correct state |\n| RESUME | Loading a prior snapshot to continue where you left off |\n\n**Why it beats `/compact`:** Compaction is lossy and exhibits recency bias. The schema forces every field — goal, decisions, completed steps, pending steps, constraints, discovered issues, modified files — to be explicit. Nothing is silently dropped.\n\n```\n/handoff                          # save to .claude/handoffs/\u003ctimestamp\u003e-\u003cslug\u003e.json\n/handoff auth-refactor.json       # save to explicit path\n/handoff load auth-refactor.json  # resume from file\n```\n\n---\n\n## Adding a New Skill\n\n1. Create a folder under `skills/` with a lowercase-hyphenated name\n2. Add a `SKILL.md` with YAML frontmatter and process instructions\n3. Add reference files in `references/` if needed\n4. Add examples in `examples/` for better activation rates\n5. Add an entry to `.claude-plugin/marketplace.json`\n6. Update this README table\n7. Open a pull request\n\n```\nskills/\u003cskill-name\u003e/\n├── SKILL.md              Required — metadata + instructions\n├── references/            Optional — detailed reference material\n└── examples/              Optional — example input/output pairs\n```\n\n---\n\n## Author\n\n**Philip A Senger**\n\n- GitHub: [@psenger](https://github.com/psenger)\n\n---\n\n## License\n\nThis project is licensed under the [MIT License](LICENSE).\n\nCopyright (c) 2026 Philip A Senger\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpsenger%2Fai-agent-skills","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fpsenger%2Fai-agent-skills","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpsenger%2Fai-agent-skills/lists"}