{
  "markdown": "![Ship Shit Dev Skills](./assets/banner.svg)\n\n# Ship Shit Dev Skills\n\n![Project Type](https://img.shields.io/badge/Project-Skills-blue)\n\n<!-- catalog-summary:start -->\n186 AI agent skills for development workflows. Works with Claude Code, OpenAI Codex, and Cursor.\n\nCatalog: **186 skills · 29 commands · 13 bundles · 199 plugins**.\n<!-- catalog-summary:end -->\n\nSkills are **model-agnostic playbooks**: the harness supplies the model, so no skill names a concrete model — orchestrators speak in capability tiers, and each repo's routing block maps tiers to models. Enforced by `scripts/validate-skill-sync.sh`; standards live in `.agents/memory/system/skill-standards.md`.\n\n## Directory Structure\n\n<!-- catalog-layout:start -->\n| Path | Role | Generated fact |\n|---|---|---|\n| `.agents/` | Repository memory, standards, and maintenance skills | Tracked |\n| `.claude/` | Claude loader adapters for shared maintenance content | Tracked |\n| `.claude-plugin/` | Generated Claude marketplace catalog | 199 generated plugins |\n| `.codex/` | Codex loader adapters for shared maintenance content | Tracked |\n| `.github/` | Issue templates and GitHub Actions workflows | Tracked |\n| `.husky/` | Git hook configuration | Tracked |\n| `.tmp/` | Tracked repository content | Tracked |\n| `assets/` | Static repository assets | Tracked |\n| `bundles/` | Generated marketplace bundle snapshots | 13 generated bundles |\n| `commands/` | Claude Code/plugin command adapters | 29 command adapters |\n| `docs/` | Human-facing orientation pages for flagship skills | Tracked |\n| `prompts/` | Shared prompt resources | Tracked |\n| `resources/` | Authoring references and supporting documentation | Tracked |\n| `scripts/` | Validation, generation, migration, and audit tooling | Tracked |\n| `skills/` | Canonical public Agent Skills sources | 186 canonical skills |\n<!-- catalog-layout:end -->\n\n## What's Included\n\n- **The Dev Loop**: a board-driven autonomous workflow that ships software with\n  `gh` + Claude + Codex.\n- **Engineering Workflows**: code review, debugging, QA, release, GitHub triage,\n  branch cleanup, production audit, and CI repair.\n- **Implementation Specialists**: frontend, backend, infrastructure, security,\n  testing, payments, and workspace setup skills.\n- **Agent Engineering**: context engineering, memory systems, evaluation,\n  multi-agent patterns, MCP building, and prompt/tool design.\n- **Maintenance Tooling**: scripts that validate skills, regenerate bundles, and\n  keep the marketplace catalog aligned with source.\n\n## The Dev Loop — ship software with gh + Claude + Codex\n\nThe flagship workflow: a **board-driven autonomous dev loop** that turns a GitHub\nissue into a reviewed PR, with you as architect/reviewer. It is the open, `gh`-driven\nversion of **ShipCode**'s pipeline — same stages, no app required.\n\n- **Plan** (human): `/ask` if you don't know which skill; `interview` → `prd-writer` / `writing-plans` write the PRD\n  and implementation plan onto the issue.\n- **Dispatch** (human opt-in): apply `dispatch:claude` (Claude lane),\n  `dispatch:codex` (Codex lane), or `dispatch:openrouter` (OpenRouter lane) to a\n  **Backlog** issue.\n- **Execute** (AI): the loop claims it → **In Progress** → branch → implement →\n  `qa-reviewer` + tests → PR, advancing `loop:*` phase labels as it goes.\n- **Review** (human): the PR lands in **Human Review**, auto-assigned to you. Merge =\n  Done; reject = back to Backlog.\n\nBoard columns are for humans (**Backlog · In Progress · Human Review · Done ·\nDeferred**); the AI loop's sub-phases ride as `loop:*` labels. Status is the GitHub\nProjects board `Status` field — the single source of truth.\n\n```bash\n# 1. Install the dev-loop bundle (every skill the loop needs)\n/plugin marketplace add shipshitdev/skills\n/plugin install shipshitdev-dev-loop@shipshitdev\n\n# 2. Provision your repo: labels + board + workflows + secrets (idempotent)\nbash scripts/setup-dev-loop.sh            # --dry-run to preview\n\n# 3. Write your repo's routing block (tracker + labels + domain). In Claude Code:\n/setup-agent-routing\n\n# 4. Drive it\n/loop                                     # Phase 1: local pull (Claude lane)\n# …or apply dispatch:claude / dispatch:codex to fire Phase 2 (GitHub Actions)\n```\n\nFull reference: [`.agents/memory/system/ai-dev-loop.md`](.agents/memory/system/ai-dev-loop.md).\n\n## Installation\n\n### Quick Install (Recommended)\n\n```bash\n# Install all skills globally for Claude Code, Codex, and Cursor\nnpx skills add shipshitdev/skills -g --agent claude-code codex cursor --skill '*' -y\n\n# Install specific skills\nnpx skills add shipshitdev/skills -g --skill stripe-implementer -y\n\n# List available skills\nnpx skills add shipshitdev/skills --list\n```\n\n> **Do NOT use `--all`** — it installs to every agent the CLI knows about (30+).\n> Always use `--agent` to target only the agents you use.\n\n### Migrating from the retired shell installer\n\nThe legacy `scripts/install-skills.sh` installer was removed. It maintained a\nsecond hardcoded bundle inventory, targeted Claude-only project paths, and could\nrecursively replace real directories during install or restore. Do not restore or\nrun an old copy.\n\nMigrate without deleting existing directories first:\n\n1. Run the supported Quick Install command above with an explicit `--agent` list.\n   The skills CLI validates requested names against the repository catalog and owns\n   the managed installation layout.\n2. Confirm the expected skills appear in each selected agent. Keep the previous\n   directories in place until that verification is complete.\n3. Review old links/directories individually. Remove only a path you have identified\n   as an obsolete managed link; never recursively delete an agent skills directory\n   as part of migration.\n\nThe README installation commands and `catalog.json` are the installation/catalog\nsource of truth. Marketplace bundles remain available through the Claude Code plugin\nalternative below.\n\n### Project-local Install\n\n```bash\nnpx skills add shipshitdev/skills --agent claude-code codex cursor\n```\n\n### Claude Code Plugin (Alternative)\n\n```bash\n/plugin marketplace add shipshitdev/skills\n/plugin install shipshitdev-dev-loop@shipshitdev\n/plugin install shipshitdev-dev-workflow@shipshitdev\n/plugin install shipshitdev-github@shipshitdev\n/plugin install shipshitdev-testing@shipshitdev\n/plugin install shipshitdev-security@shipshitdev\n```\n\n### Platform setup boundary\n\nThe portable workflow source is always `skills/<name>/SKILL.md`. Platform entry\nsurfaces should route to that source instead of copying it.\n\n- **Codex project instructions:** use `AGENTS.md` and optional nested\n  `AGENTS.override.md` files. Global preferences live in `~/.codex/AGENTS.md`.\n- **Codex configuration:** user defaults live in `~/.codex/config.toml`; trusted\n  repositories may add `.codex/config.toml` overrides. Model, effort, approvals,\n  sandboxing, workspace, and other execution settings belong there or in the app.\n- **Codex workflows:** install shared skills under `.agents/skills`. Do not add `.codex/instructions.md` or `.codex/commands`;\n  neither is a supported project instruction or reusable workflow surface.\n- **Deprecated Codex prompts:** `~/.codex/prompts` is user-local, explicit-only,\n  and deprecated in favor of skills. This repository does not publish prompts\n  there.\n- **Claude commands:** files under `commands/` are thin Claude Code/plugin\n  conveniences. They are not included in generated bundle manifests and are not\n  presented as Codex commands unless a supported adapter is added later.\n\n### For Contributors\n\n```bash\ngit clone https://github.com/shipshitdev/skills.git ~/shipshitdev-skills\ncd ~/shipshitdev-skills\nnpx skills add . -g --agent claude-code codex cursor --skill '*' -y\n```\n\n## Adding Skills & Commands\n\n### Adding a Skill\n\n1. Create directory in `skills/skill-name/`\n2. Add `SKILL.md` with YAML frontmatter\n3. Update this README\n\n```bash\nmkdir -p skills/my-skill\ntouch skills/my-skill/SKILL.md\n```\n\n### Adding a Command\n\nCommands are Claude Code/plugin entry points, not portable workflow definitions.\nRoute each command to a canonical skill whenever the workflow should work in Codex\nor another Agent Skills-compatible harness.\n\n1. Create a thin `.md` file in `commands/`\n2. Follow naming: `{verb}-{noun}.md`\n3. Update this README\n\n## Documentation\n\n- `.agents/memory/system/skill-standards.md` - Agent Skills spec + Claude Code extensions\n- `.agents/memory/system/skill-management.md` - Single-source skill workflow\n- `.agents/memory/system/architecture.md` - .agents folder structure\n- `.agents/memory/system/platform-adaptations.md` - Claude vs Codex writing guide\n- `.agents/memory/system/execution-boundary.md` - app-owned execution settings vs reusable content\n- `.agents/memory/system/routine-standards.md` - portable scheduled-task and automation authoring\n- `.agents/memory/system/ai-dev-loop.md` - Board-driven autonomous dev loop (`/loop` + agent-dispatch)\n- `docs/skills/dev-loop.md` - Human-facing Dev Loop orientation\n- `docs/skills/ask-dev-loop.md` - Human-facing router for the flagship flow\n\n## Development Commands\n\n| Command | Description |\n|---------|-------------|\n| address | Resolve PR review comments — propose fixes and replies |\n| agent | Audit, configure, scaffold, and route agents and subagents |\n| ask | Name the Dev Loop skill that fits the current situation |\n| board | Set up, sync, schedule, and review a GitHub Projects v2 board |\n| bug | File a GitHub bug issue |\n| cleanup | Prune merged branches, stale worktrees, and finished work |\n| codex-loop | Claim and work one dispatch:codex issue locally via codex exec |\n| deploy | Deploy the app and provision infrastructure |\n| design | Review and refine UI — audit, critique, polish, layout |\n| deslop | Remove AI slop and tells from code — shortcut for `/refactor deslop` |\n| env | Discover, scaffold, and validate environment variables via `env-setup` |\n| feature | Capture requirements into PRD epics and GitHub issues |\n| fix-ci | Diagnose and fix failing CI checks on a PR |\n| loop | Claim and work one dispatch issue end-to-end |\n| merge | Merge all approved open PRs into the trunk in one gated sweep |\n| pr | Create, update, and publish a pull request |\n| prd | Create specs, PRDs, and feature plans |\n| prompt | Optimize a prompt via `prompt-engineering`'s 4-D workflow |\n| qa | Run a structured verification pass before commit — shortcut for `/test qa` |\n| refactor | Improve existing code — deslop, refactor, debt, perf, structure, stack |\n| release | Cut a trunk release with patch notes |\n| review | Review changes, a PR, all open PRs, or recent commits — natively or via the Grok CLI |\n| roadmap | Turn ICP into a revenue-ranked backlog and dated milestones |\n| scan | Run a security audit — full app/API audit or dependency supply chain |\n| skill | Author, capture, test, and scout agent skills |\n| standup | Summarize what you shipped over a time window |\n| suggest | Post inline suggested changes on a PR |\n| test | One testing front door — run, qa, tdd, e2e, coverage, init, regression |\n| wait-what | Re-pitch the last message in plain English |\n\n### Intentional shortcuts, not duplicates\n\n- `/qa` = `/test qa` and `/deslop` = `/refactor deslop` — the two most-used\n  modes keep their own top-level spelling on purpose.\n- `/address`, `/suggest`, and `/fix-ci` are three distinct jobs in the same PR\n  flow, not overlapping reviews: `/review` finds issues *for you*, `/suggest`\n  posts suggestions *onto someone else's PR*, `/address` resolves comments\n  *others left on yours*, and `/fix-ci` repairs the failing checks. `/merge`\n  then lands what's green.\n- `/performance` and `/tests` were retired: performance analysis lives at\n  `/refactor perf`, and every `/tests <scope>` spelling is now\n  `/test run <scope>`.\n\n<!-- catalog-skills-heading:start -->\n## Skills (186)\n<!-- catalog-skills-heading:end -->\n\n### Dev Loop (15)\n\n`interview`, `grilling`, `domain-modeling`, `ask-dev-loop`, `wait-what`, `feature-intake`, `prd-writer`, `writing-plans`, `prd-quality-gate`, `prd-task-creator`, `executing-plans`, `setup-agent-routing`, `gh-project-board`, `gh-board-sync`, `qa-reviewer`\n\n### Dev Workflow (43)\n\n`agent-architecture-audit`, `agent-config-audit`, `ai-agent-cost-optimizer`, `ai-regression-testing`, `codebase-advisor`, `codebase-design`, `code-review`, `structural-review`, `full-code-review`, `grok-review`, `review-dispatch`, `commit-summary`, `changelog-generator`, `standup`, `deslop`, `refactor-dispatch`, `tech-debt`, `stack-modernization`, `debug`, `deploy`, `execution-debugging`, `deployment-composer`, `docs`, `llm-structured-output`, `merge-open-prs`, `production-audit`, `refactor-code`, `release`, `release-dispatch`, `git-cleanup`, `release-pr-gates`, `scaffold`, `shape`, `skill-capture`, `skill-comply`, `skill-scout`, `systematic-debugging`, `receiving-code-review`, `verification-before-completion`, `wizard`, `worktree`, `skill-dispatch`, `deploy-dispatch`\n\n### GitHub (20)\n\n`bug`, `gh-address-comments`, `gh-board-sync`, `gh-fix-ci`, `gh-inbox`, `gh-pr-publish`, `gh-project-board`, `gh-review-suggestions`, `github-actions-author`, `git-safety`, `feature-intake`, `merge-open-prs`, `release`, `release-dispatch`, `git-cleanup`, `release-pr-gates`, `worktree`, `finishing-a-development-branch`, `pr-comments`, `fix-merge-conflicts`\n\n### Testing (9)\n\n`playwright-e2e-init`, `tdd`, `testing-expert`, `testing-cicd-init`, `qa-reviewer`, `qa-loop`, `husky-test-coverage`, `test-runner`, `test-dispatch`\n\n### Frontend & React (32)\n\n`frontend-design`, `component-library`, `accessibility`, `audit`, `clarify`, `critique`, `design-consistency-auditor`, `html-style`, `layout`, `polish`, `quieter`, `react-component-performance`, `react-hook-form`, `theme-factory`, `prototype`, `react-patterns`, `react-refactor`, `react-testing-library`, `react-native-components`, `expo-architect`, `landing-page-vercel`, `micro-landing-builder`, `ai-loading-ux`, `table-filters`, `quick-view`, `nextjs-validator`, `nextra-writer`, `shadcn`, `shadcn-setup`, `tailwind`, `tailwind-validator`, `design-dispatch`\n\n### Backend & Data (9)\n\n`api-design-expert`, `error-handling-expert`, `graphql-architect`, `nestjs-expert`, `nestjs-testing-expert`, `incremental-fetch`, `turborepo`, `typescript-expert`, `typescript-refactor`\n\n### Infrastructure (12)\n\n`docker-expert`, `aws-infrastructure`, `ec2-backend-deployer`, `vercel-deploy`, `postgres-ops`, `mongodb-migration-expert`, `mongodb-atlas-checker`, `monitoring-setup`, `nestjs-queue-architect`, `performance-expert`, `redis-caching`, `security-expert`\n\n### Security (5)\n\n`security-audit`, `security-expert`, `git-safety`, `dependency-audit`, `open-source-checker`\n\n### AI Agents (17)\n\n`prompt-engineering`, `mcp-builder`, `skill-creator`, `context-fundamentals`, `context-optimization`, `context-degradation`, `context-engineering`, `memory-systems`, `multi-agent-patterns`, `tool-design`, `evaluation`, `advanced-evaluation`, `comment-mode`, `spec-first`, `agent-browser`, `agent-dispatch`, `codex-image-gen`\n\n### Workspace Setup (13)\n\n`fullstack-workspace-init`, `project-init-orchestrator`, `env-setup`, `linter-formatter-init`, `clerk-validator`, `content-script-developer`, `package-architect`, `artifacts-builder`, `wizard`, `open-source-checker`, `devcontainer-setup`, `biome-validator`, `bun-validator`\n\n### Planning & PRDs (15)\n\n`icp`, `roadmap-analyzer`, `roadmap-to-milestones`, `cto-advisor`, `interview`, `grilling`, `domain-modeling`, `prototype`, `feature-intake`, `prd-task-creator`, `prd-writer`, `prd-quality-gate`, `rules-capture`, `writing-plans`, `prd-dispatch`\n\n### Payments & Product Integrations (1)\n\n`stripe-implementer`\n\n### Session Management (5)\n\n`agent-folder-init`, `executing-plans`, `setup-agent-routing`, `wait-what`, `workspace-performance-audit`\n\n## How Skills Adapt to Projects\n\nSkills are **adaptive** - they scan project documentation to understand:\n\n- Project architecture and structure\n- Tooling, scripts, and CI conventions\n- Existing patterns and conventions\n- Domain terminology and implementation constraints\n\nIf a project has its own skill, the generic skill will collaborate with or defer to it.\n\n## Publishing & CI/CD\n\nWhen you push to `master`, GitHub Actions automatically regenerates the `bundles/` directory to keep marketplace plugins in sync with skills.\n\n### Claude Marketplace\n\nUsers install directly from GitHub:\n\n```bash\n# Add the marketplace\n/plugin marketplace add shipshitdev/skills\n\n# Install category bundles\n/plugin install shipshitdev-dev-loop@shipshitdev\n/plugin install shipshitdev-dev-workflow@shipshitdev\n/plugin install shipshitdev-github@shipshitdev\n/plugin install shipshitdev-testing@shipshitdev\n/plugin install shipshitdev-frontend@shipshitdev\n/plugin install shipshitdev-backend@shipshitdev\n/plugin install shipshitdev-infrastructure@shipshitdev\n/plugin install shipshitdev-security@shipshitdev\n/plugin install shipshitdev-session@shipshitdev\n/plugin install shipshitdev-planning@shipshitdev\n```\n",
  "bytes": 16988,
  "sha": "bafc24f939c0136de37c163098ed746c1540ed85bed9e65aba0f3ba91bc72140",
  "repo_slug": "shipshitdev/skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/skl_shipshitdev_skills_youtube_video_analyst_b8b511b5/readme"
}