Skip to content AI News: Nano Banana 2.1 is out: better edits and text, half the API price per image →
SK-31 · Skill · code · rev. 10/2026

Addy Osmani's agent skills for Claude Code

agent-skills is Addy Osmani's pack of 25 engineering skills for coding agents, organized as a development lifecycle: define, plan, build, verify, review and ship, with a slash command for each stage.

Publisheraddyosmani Community
LicenseMIT
Works inClaude CodeCodex
Checked7 Oct 2026
Install
/plugin install agent-skills@addy-agent-skills
Source on GitHub ↗★ 103K 7 Oct 2026

Addy Osmani's pack writes down the workflows and quality gates a senior engineer follows, so an agent applies them the same way every time. It holds 25 skills, 24 for stages of the lifecycle plus one that maps a request to the right skill, and each comes with steps, checks and a table of excuses the agent should not accept from itself. The license is MIT. For the basics, see Claude Code skills.

The nine commands

CommandStagePrinciple, per the README
/specDefine what to buildSpec before code
/planPlan howSmall, atomic tasks
/buildBuild in slicesOne slice at a time
/testProve it worksTests are proof
/constraintsSet the quality barDecide it once, enforce it everywhere
/reviewReview before mergeImprove code health
/webperfAudit web performanceMeasure before you optimize
/code-simplifySimplifyClarity over cleverness
/shipReleaseFaster is safer

/build auto writes the plan from an approved spec and implements every task in one pass after you approve the plan once. Skills also trigger on their own: designing an API brings in the API design skill, building UI the front-end one.

Skills worth knowing

  • interview-me: asks one question at a time until it is about 95% sure what you want.
  • spec-driven-development: a short requirements document before any code.
  • incremental-implementation: thin vertical slices, each tested and committed.
  • source-driven-development: grounds framework decisions in official documentation and flags what it could not verify.
  • doubt-driven-development: a fresh-context, adversarial review of each important decision while the work runs.
  • code-review-and-quality: reviews on five axes, with changes kept to about 100 lines.
  • security-and-hardening, performance-optimization, ci-cd-and-automation and deprecation-and-migration for the later stages.

How to install it

Claude Code, from the author's marketplace:

/plugin marketplace add addyosmani/agent-skills
/plugin install agent-skills@addy-agent-skills

If the install fails with an SSH permission error, add the marketplace by its full HTTPS address, https://github.com/addyosmani/agent-skills.git. For any of 70-plus agents, or to pick single skills:

npx skills add addyosmani/agent-skills
npx skills add addyosmani/agent-skills --skill interview-me

A single-skill install copies only that skill's folder, so links to the shared checklists in the repo's references/ folder will not resolve. Install the whole pack if you want them.

When to use it

  • Teams that want one shared definition of "done" for agent-written code.
  • New projects, where the full lifecycle from /spec to /ship fits. On an existing codebase the author's adoption guide suggests starting with the verification skills.

Superpowers covers similar ground with a stricter fixed order. Pick one; two process packs at once will compete for the same steps. If you only want the questioning step, grill-me is one skill instead of a pack.

Sources

  1. addyosmani/agent-skills README — GitHub, accessed October 2026
Weekly

New prompts in your inbox

One email a week: the best new prompts and one short guide. Unsubscribe any time. Privacy policy