Skills addyosmani agent-skills
Addy's Agent Skills Bundle of 25 skills
- Use it when:
- You want your agent to follow a senior engineer's process at every phase, from a vague idea to a production launch.
- It does:
- Gives each phase its own skills: interview, spec, plan, build in tested slices, debug, review, simplify, secure, commit, document and ship.
- You get:
- A spec you approve, a plan in tasks/plan.md and tasks/todo.md, small tested commits, a five-axis code review and a launch checklist with a rollback plan.
- Runs:
- By itself: each skill loads when the work matches it, and using-agent-skills maps a task to its skill when the choice isn't obvious.
- Commands:
- In Claude Code, slash commands such as /spec, /plan, /build, /test, /review and /ship start the matching skills; /build auto works through a whole plan after one approval.
- Works with:
- Claude Code, Codex, Gemini CLI, Antigravity, Command Code and any agent the skills CLI supports.
- Install whole:
- A single-skill install copies only that skill's folder, so its links to the repo's shared checklists break, though the skill still works.
TL;DR
Addy Osmani's pack maps software work onto six phases. using-agent-skills, the entry point, picks the skill for a task and sets shared rules: state assumptions, push back, stay in scope, verify before claiming done. Define: interview-me finds out what you actually want, idea-refine turns a rough idea into a one-pager, spec-driven-development writes the spec and waits for your approval, and constraint-driven-development records the quality bar in CONSTRAINTS.md. Plan: planning-and-task-breakdown turns the spec into ordered tasks. Build: context-engineering loads the right context and source-driven-development checks framework code against official docs; incremental-implementation and test-driven-development then deliver each task as small tested commits, helped by frontend-ui-engineering and api-and-interface-design, with observability-and-instrumentation adding telemetry alongside and doubt-driven-development sending non-trivial decisions to an adversarial reviewer. Verify: browser-testing-with-devtools and debugging-and-error-recovery. Review: code-review-and-quality, code-simplification, security-and-hardening and performance-optimization. Ship: git-workflow-and-versioning, ci-cd-and-automation, documentation-and-adrs, deprecation-and-migration and shipping-and-launch. Every skill but the router has a table of excuses to reject and a verification checklist, and works alone. The repo also ships reviewer personas and shared checklists that only a whole-repo install brings.
Summarizes the repo at 301b02b.
Any agent, with the skills CLI
npx skills add addyosmani/agent-skills Claude Code plugin
claude plugin marketplace add addyosmani/agent-skills && claude plugin install agent-skills@addy-agent-skills Codex plugin
codex plugin marketplace add addyosmani/agent-skills && codex plugin add agent-skills@agent-skills Gemini CLI
gemini skills install https://github.com/addyosmani/agent-skills.git --path skills Antigravity CLI plugin
agy plugin install https://github.com/addyosmani/agent-skills.git Command Code, which asks which skills to install
cmd skills add addyosmani/agent-skills Source on GitHubAll skills from addyosmani/agent-skills
- Total Downloads
- 83,532 of code-review-and-quality, its most installed skill, on skills.sh, Nov 9, 2026
- Repo stars
- 114,200 addyosmani/agent-skills
- First seen
- Feb 16, 2026 on skills.sh, its first skill
Its skills, in the order they're used
The ones marked “Only in the bundle” need the rest of it. The others also work alone.
-
Routes each piece of work to the right skill of Addy Osmani's pack, and sets rules every skill shares.
Routes each task to the bundle's other skills; alone, it only adds its shared rules.
-
interview-me
65.9K total downloadsInterviews you one question at a time, each with its best guess attached, until it can predict your answers, then confirms what you want.
-
idea-refine
67.9K total downloadsTurns a vague idea into a one-page proposal by expanding it into variations, stress-testing the best, and naming what not to build.
-
spec-driven-development
74.1K total downloadsWrites a spec covering objective, commands, structure, code style, testing and boundaries, and stops for your approval before planning or code.
-
constraint-driven-development
41.9K total downloadsInterviews you for a quality bar, records it in CONSTRAINTS.md with the commands that enforce it, and catches diffs that weaken it.
-
planning-and-task-breakdown
73K total downloadsBreaks a spec into small, ordered tasks, each with acceptance criteria and a check, in a plan file and task list.
-
context-engineering
70.6K total downloadsSets up what the agent sees: a rules file, the right spec and source files per task, and trimmed context in long sessions.
-
source-driven-development
66.9K total downloadsChecks framework-specific code against the official docs for your exact versions, and cites the source for every pattern it uses.
-
incremental-implementation
70.5K total downloadsBuilds a change in thin slices, each implemented, tested, verified and committed before the next, touching only what the task needs.
-
frontend-ui-engineering
77.3K total downloadsShapes UI code: composable components, simple state, your design system's tokens, WCAG 2.1 AA accessibility and no generic AI look.
-
api-and-interface-design
72K total downloadsDesigns APIs and module interfaces that are hard to misuse: contract first, one error format, validation at the edges, changes that don't break callers.
-
observability-and-instrumentation
63.3K total downloadsInstruments a feature with structured logs, metrics, traces and symptom-based alerts, each tied to a question on-call will ask.
-
doubt-driven-development
65.3K total downloadsSends each non-trivial decision to a fresh-context reviewer told to find what's wrong, then weighs every finding before the decision stands.
-
test-driven-development
70.8K total downloadsWrites a failing test before the code that makes it pass, and reproduces every bug as a test before fixing it.
-
browser-testing-with-devtools
67.3K total downloadsChecks and debugs web pages in a real Chrome through the Chrome DevTools MCP server: DOM, console, network, performance and screenshots.
-
debugging-and-error-recovery
71.9K total downloadsStops feature work when something breaks, then reproduces, localizes, reduces and fixes the root cause, guarded by a regression test.
-
code-review-and-quality
83.5K total downloadsReviews a change on five axes, correctness, readability, architecture, security and performance, with every finding labelled by severity.
-
code-simplification
72.1K total downloadsSimplifies working code for readability while keeping its behavior exactly the same, one tested change at a time.
-
security-and-hardening
80.8K total downloadsThreat-models a feature, then hardens it against injection, broken access control, leaked secrets, risky dependencies, privacy gaps and LLM prompt injection.
-
performance-optimization
74.6K total downloadsMeasures before optimizing: finds the real bottleneck, fixes only that, re-measures, and reverts any change that doesn't beat the noise.
-
git-workflow-and-versioning
68.3K total downloadsKeeps git history clean: short-lived branches, small atomic commits with messages that explain why, and releases tagged with semantic versions and changelogs.
-
ci-cd-and-automation
66.8K total downloadsSets up CI that blocks merges until lint, type checks, tests, build and a security audit pass, and plans deploys you can roll back.
-
documentation-and-adrs
73K total downloadsRecords why decisions were made in architecture decision records, and sets rules for comments, API docs, READMEs and changelogs.
-
deprecation-and-migration
65.9K total downloadsRetires old systems, APIs or features safely: replacement first, consumers migrated one by one, old code removed once nothing uses it.
-
shipping-and-launch
67K total downloadsPrepares a production launch: a pre-launch checklist, feature flags, a staged rollout with rollback thresholds, monitoring and a written rollback plan.