A collection of skills for AI coding agents. Skills are packaged instructions and workflows that extend agent capabilities across any tool that supports the Agent Skills format.
Compatible with Devin, Claude Code, Cursor, Codex, Gemini CLI, Windsurf, GitHub Copilot, Cline, Roo, and many more.
Install all skills:
npx skills@latest add amit-t/skillsInstall a specific skill:
npx skills@latest add amit-t/skills --skill tdd
npx skills@latest add amit-t/skills --skill prd-draft
npx skills@latest add amit-t/skills --skill eng-specInstall for a specific agent:
npx skills@latest add amit-t/skills --agent claude-code cursorInstall globally (available in all projects):
npx skills@latest add amit-t/skills -g| Skill | Description |
|---|---|
prd-draft |
Create modern, AI-era PRDs through guided interview |
prd-review-panel |
Multi-agent PRD review from 7 perspectives |
prd-approve |
Approve a reviewed PRD and update pipeline |
prd-to-plan |
Turn a PRD into a phased implementation plan using tracer-bullet vertical slices |
write-a-prd |
Create a PRD through interview, codebase exploration, and module design |
| Skill | Description |
|---|---|
pmo-status |
Full project status across Product, Engineering DOE, and Implementation |
| Skill | Description |
|---|---|
code-review-multi-axis |
Slash-only principal-engineer review in two modes — pre-PR: Standards + Spec axes side by side vs a fixed point (nothing posted); post-PR: two-phase GitHub review walking 11-dimension findings one at a time before posting. Formerly code-review |
docs-from-prs |
Survey merged PRs, fill doc gaps in README and other user-facing docs, audit drift hot spots, and run a grammar/casing/alignment pass |
domain-grill |
Engineering-only stress-test of an eng spec / TDD / refactor / tech design against CONTEXT.md — pick a depth (quick / standard / deep, default deep) at start; use /grill-me for PRDs and non-technical plans |
e2e-audit |
Playwright PRD audit + persona screenshot demo + auto-emit one ralph task per bug to fix_plan.md |
eng-spec |
Convert approved PRD into TDD + Spec + ADRs with 5-agent review |
gh-pages-neo-brutalist |
Drop-in Jekyll templates that scaffold a GitHub Pages site with a neo-brutalist design system (4 themes, IBM Plex Mono, hard borders, offset shadows) |
gh-repo-mirror |
Scaffold a new GitHub repo that mirrors a reference repo's settings, branch protection, security flags, team + direct-collaborator access, and (optional) GitHub Pages site |
git-guardrails-claude-code |
Block dangerous git commands before they execute |
grill-me |
Stress-test a plan or design through relentless interview — pick a depth (quick / standard / deep, default deep) at start |
grill-me-auto |
Batch-mode grill-me: write every question, option, recommendation, and alt into one collapsible markdown grill document for one-shot replies |
package-scout |
Research, compare, and select the best packages before installing any dependency |
qa |
Interactive QA session: report bugs conversationally, agent files GitHub issues |
repo-context-scan |
Scan a codebase to build CONTEXT.md (or CONTEXT-MAP.md for multi-context repos) and seed ADRs for clearly-deliberate decisions |
request-refactor-plan |
Create a detailed refactor plan with tiny commits via interview, filed as GitHub issue |
resume-tailoring |
Tailor a resume to a specific job — research, branching discovery, confidence-scored matching, MD+DOCX+PDF+report |
setup-amit-skills |
Configure a repo for amit-t/skills conventions via a guided 3-decision setup (agent instruction surface, context layout, catalog/category mode) — modeled on Matt Pocock's setup-matt-pocock-skills |
setup-pre-commit |
Set up Husky + lint-staged + Prettier pre-commit hooks (typecheck + test) in a JS/TS repo — ported from Matt Pocock's misc/setup-pre-commit |
tdd |
Test-driven development via the red-green loop — seams, anti-patterns, and vertical slices; refactoring moved to two-axis-review |
two-axis-review |
Review a diff since a fixed point along two parallel axes — repo standards (incl. a Fowler code-smell baseline) and spec fidelity — reported side by side; ported from Matt Pocock's engineering/code-review |
ubiquitous-language |
Deprecated — superseded by /repo-context-scan (build) and /domain-grill (stress-test) |
| Skill | Description |
|---|---|
design-interview |
Interactive design brief interview before generating screens |
design-draft |
Full UXD workflow from design interview to developer handoff |
design-review |
Multi-agent design review from 5 perspectives |
| Skill | Description |
|---|---|
leadership-update |
Reformat raw notes into an outcome-first, three-sentence update with a clear ask |
wisdom-capture |
Capture a wisdom snippet into a personal corpus — categorize into a closed bucket, write wisdoms/<YYYY>/<MM>/<ulid>.md with frontmatter, commit |
| Skill | Description |
|---|---|
compact-conversation |
Compact the current conversation into a concise summary to reduce context window usage |
concise-reporting |
Ultra-concise status/progress reporting; full verbosity preserved for written artifacts |
session-handoff |
Compact the current conversation into a discoverable handoff document at .claude/handoffs/ so a fresh agent can pick it up with /resume-session-handoff (named session-handoff to avoid colliding with Devin's built-in /handoff) |
precision-mode |
Universal conciseness directive — makes every LLM response shorter, denser, and more precise |
resume-session-handoff |
Pick up where a previous session left off — loads the newest open handoff written by /session-handoff, runs an environment preflight, marks the file resumed on confirm, and asks what to do next (no auto-execute). Named resume-session-handoff to avoid shadowing agents' built-in /resume and Devin's built-in /handoff |
session-feedback |
Mine the current conversation for every correction, preference, and do-differently lesson; write a dated feedback file the next session reloads as memory |
skill-sync |
Sync an existing skill from a source path, or scaffold a new one via claude/codex/devin |
| Skill | Description |
|---|---|
write-a-skill |
Create new agent skills with proper structure |
Skills are self-contained units of functionality that teach AI coding agents how to perform specific tasks. Each skill is a SKILL.md file (with optional companion files) that gets loaded into the agent's context when invoked.
Skills can be invoked by name (e.g., /tdd, /prd-draft) or automatically triggered by the agent when relevant.
If you prefer not to use the CLI, copy skill directories into your agent's skills folder:
| Agent | Project Path | Global Path |
|---|---|---|
| Devin / Windsurf | .cognition/skills/ or .windsurf/skills/ |
~/.config/cognition/skills/ |
| Claude Code | .claude/skills/ |
~/.claude/skills/ |
| Cursor | .cursor/skills/ |
— |
| Codex | Append to AGENTS.md |
— |
| Gemini CLI | Append to GEMINI.md |
— |
MIT