PlugWiki

Using Agent Skills for Claude Code: pick the right skill for the task

The guide skill for the agent-skills pack: it makes your agent pick the right workflow skill and follow shared working rules.

Installs
63.4K
on skills.sh
Stars
112.6K
on GitHub
Updated
Nov 2026
last repo commit

using-agent-skills is the meta-skill of the agent-skills pack. It loads at the start of a session or when the agent must decide which skill applies, then points to one by phase: define, plan, build, verify, review or ship.

It also sets six operating behaviors for every skill: state assumptions, stop on confusion, push back when warranted, keep things simple, stay in scope and verify before calling work done. It routes to other skills in the same repo, so install the whole pack and not this skill alone.

Good for

  • Developers installing the whole agent-skills pack
  • Teams that want agents to state assumptions and push back
  • Anyone unsure which skill fits the task

Install using-agent-skills

Claude Code

  1. 1

    Copy this prompt.

    Install the using-agent-skills skill in Claude Code. Run this command:

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

Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Pick one route, not both, or you get every skill twice. This skill routes to other skills in the same repo, so install the whole pack. In Claude Code the README says the plugin also adds 9 slash commands such as /spec and /build. Where a built-in command has the same name, as /review does, use /agent-skills:review. If the marketplace step fails over SSH, the README suggests the full HTTPS URL: /plugin marketplace add https://github.com/addyosmani/agent-skills.git

From the project README. Source.

What's in the skill

When it loads

Discovers and invokes agent skills. Use when starting a session, or when you need to decide which skill or workflow applies to the piece of work at hand. This is the meta-skill that governs how all other skills are discovered and invoked.

Sections in SKILL.md

  1. Overview
  2. Skill Discovery
  3. Core Operating Behaviors
  4. 1. Surface Assumptions
  5. 2. Manage Confusion Actively
  6. 3. Push Back When Warranted
  7. 4. Enforce Simplicity
  8. 5. Maintain Scope Discipline
  9. 6. Verify, Don't Assume
  10. Failure Modes to Avoid
  11. Skill Rules
  12. Lifecycle Sequence
  13. Quick Reference

Key facts

Author
addyosmani (repo owner)
Installs on skills.sh
63,420 (Nov 2, 2026)
GitHub stars
112,555 on the agent-skills repo (Nov 2, 2026)
License
MIT
Last commit
2026-11-01
What installs
1 file: SKILL.md
Needs
The other skills in the agent-skills pack that it routes to; install the whole set
Works in
Claude Code, Codex, Cursor, Gemini CLI, Copilot, OpenCode and more, per the README

What it includes

  • Maps the phase of work to the skill that applies

    Define, plan, build, verify, review and ship each have named skills.

    Skill discovery

  • State assumptions before anything non-trivial

    Do not silently fill in vague requirements.

    1. Surface assumptions

  • Stop, name the confusion, ask, wait

    No guessing between conflicting requirements.

    2. Manage confusion

  • Point out problems, quantify the downside, propose an alternative

    Accept the human's decision once they have full information.

    3. Push back when warranted

  • Prefer the boring, obvious solution

    Ask whether fewer lines would do.

    4. Enforce simplicity

  • Touch only what you were asked to touch

    No unrequested cleanups or removals.

    5. Maintain scope discipline

  • Work is not done until verification passes

    Points to a Definition of Done file in the repo-level references folder.

    Read · 6. Verify, do not assume

  • Check for a skill first; follow steps in order; start with a spec when in doubt

    A 16-step sequence from interview-me to shipping-and-launch; not every task needs every step.

    Skill rules and lifecycle sequence

  • One line for each skill in the pack

    23 skills listed by phase.

    Quick reference table

  • Ten subtle errors that look like productivity

    For example wrong assumptions left unchecked, plowing ahead when lost, being agreeable about a flawed approach, building without a spec and skipping verification because it looks right.

    Failure modes to avoid

From the official docs

Permissions

Low risk

It only routes between skills. The plugin route adds the whole pack and its slash commands, so read what you enable.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • The plugin route adds the whole pack

    The README says the Claude Code plugin adds all 25 skills and 9 slash commands such as /spec and /build. The manifest (.claude-plugin/plugin.json) lists commands and skills and has no hooks entry. Use the skills CLI route if you want skills only.

  • Hook scripts in the repo are not switched on by the plugin

    hooks/session-start.sh says the plugin does not wire it, and hooks/SDD-CACHE.md tells you to add its hooks to your own settings.json. If you add any of them, read the script first.

Questions

What does using-agent-skills do?

It tells Claude how to find and apply the other skills in the agent-skills pack, and sets six working rules such as stating assumptions, stopping on confusion and verifying before finishing.

How do I install using-agent-skills?

Use the install box above, which installs the whole pack. The SKILL.md routes to other skills such as idea-refine, spec-driven-development, incremental-implementation and shipping-and-launch, so on its own it has little to route to.

Does it work without the other skills?

The rules still apply, but its discovery and lifecycle sections name other skills in the pack, and a one-skill install does not include them. The install box installs the whole set.

Does it run anything?

No commands. The SKILL.md defines routing and behavior rules and bundles no scripts.

What are the six rules?

Surface assumptions, manage confusion actively, push back when warranted, enforce simplicity, maintain scope discipline, and verify instead of assuming.

Does it work in Codex, Cursor or other tools?

The README documents Codex (codex plugin marketplace add addyosmani/agent-skills, then codex plugin add agent-skills@agent-skills, Codex CLI v0.122 or newer), Cursor (copy skills into .cursor/skills/), Gemini CLI, Windsurf, OpenCode, GitHub Copilot, Kiro and others. It says the skills are plain Markdown that work with any agent that accepts instruction files. It does not mention the Claude app.

Information

updated Nov 2, 2026

Repository
addyosmani/agent-skills
Folder
skills/using-agent-skills

addyosmani/agent-skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5
  1. agent-skills README · github.com
  2. using-agent-skills SKILL.md · github.com
  3. plugin.json · github.com
  4. hooks/session-start.sh · github.com
  5. hooks/SDD-CACHE.md · github.com

Updated Nov 2, 2026.