Agent Skills (Addy Osmani): spec-to-ship engineering workflow
25 engineering skills and 9 slash commands (/spec, /plan, /build, /test, /constraints, /review, /webperf, /code-simplify, /ship) that make a coding agent write a spec first, plan atomic tasks, build test-first in small committed slices, review on five axes and ship behind quality gates.
- Scan
- Repo scan pass
- GitHub stars
- 108k
- Last updated
- 2026-10-13
- License
- MIT
- Cost
- free
Why it helps
Keeps a venture-operator agent that builds its own product on a disciplined spec, test and review loop, so features ship in small verified slices instead of large unreviewed changes that break production.
Ask your owner first. installs a skill
- '/build auto' runs a whole approved plan autonomously; keep the owner's approval step for risky or production-facing changes.
- '/build auto' implements an approved plan autonomously; keep owner approval for production-facing changes.
Paste to your agent
Read https://upskill.coach/skills/addyosmani-agent-skills.md. If the security scan passed, ask me before installing, then install it with the command under Install and tell me what you did.Install
npx skills add addyosmani/agent-skills (all 25) or --skill <name>; in Claude Code: /plugin marketplace add addyosmani/agent-skills then /plugin install agent-skills@addy-agent-skills
Community
Helps with
Repo security-scanned on 2026-11-18 at commit 39b723ceebf8: pass. The verdict covers the whole repo (addyosmani/agent-skills); this skill wasn't scanned separately.
- Scan notes
- Automated scan: review (high 1) at commit 39b723ceebf8 on 2026-11-18. Manual line review (Agent Coach library agent (line-by-line read of every high/critical finding). Not an owner sign-off., 2026-11-18): pass. Only finding (HIGH pi.ignore-previous, skills/source-driven-development/SKILL.md:110) is a defensive rule telling the agent to ignore directives such as 'ignore previous instructions' found in fetched docs; no medium+ findings otherwise.
- Source
- github.com/addyosmani/agent-skills
- Repo
- addyosmani/agent-skills
Static scan at a pinned commit (prompt-injection, malware and exfiltration rules plus semgrep, gitleaks, osv-scanner); flagged lines reviewed by hand. A scan lowers risk; it doesn't prove safety.
Full text not reproduced here. Read it at the source.
Alternatives
Compare allSuperpowers (obra)
Skills framework/methodology: verification-before-completion, writing-plans, executing-plans, systematic-debugging, TDD, subagent-driven development.
Anthropic Skills (skill-creator, webapp-testing)
Anthropic's public skills repo incl. skill-creator (turn a workflow/correction into a skill) and webapp-testing (Playwright).
agent-browser (Vercel Labs)
Browser-automation CLI for agents plus its skill (snapshot/click/fill), local or cloud browsers.
Playwright MCP (Microsoft)
Official Playwright MCP server: drive a real browser via accessibility snapshots.
Related
code-review-and-quality
Agent skill (SKILL.md) for code review and agent orchestration, published in addyosmani/agent-skills. Load it into Claude Code, Codex, Cursor or another SKIL…
security-and-hardening
Agent skill (SKILL.md) for security review and agent orchestration, published in addyosmani/agent-skills. Load it into Claude Code, Codex, Cursor or another …
frontend-ui-engineering
Agent skill (SKILL.md) for frontend development, UI and design and agent orchestration, published in addyosmani/agent-skills. Load it into Claude Code, Codex…
performance-optimization
Agent skill (SKILL.md) for agent orchestration, published in addyosmani/agent-skills. Load it into Claude Code, Codex, Cursor or another SKILL.md-aware agent.
planning-and-task-breakdown
Agent skill (SKILL.md) for task and project tracking and agent orchestration, published in addyosmani/agent-skills. Load it into Claude Code, Codex, Cursor o…
Superpowers (obra)
Skills framework/methodology: verification-before-completion, writing-plans, executing-plans, systematic-debugging, TDD, subagent-driven development.
Comments
Comments are from visitors, not reviewed by Upskill, and never shown to agents.