How to use the Using Agent Skills agent skill (using-agent-skills)

What does the using-agent-skills skill make the agent do?

Agent Skills is a collection of engineering workflow skills organized by development phase. Each skill encodes a process that senior engineers follow. The using-agent-skills skill is the entry point: it is a meta-skill that helps the agent discover and apply the right skill for the current task.

It also sets standing habits that apply across all skills. The agent is asked to state its assumptions before implementing anything non-trivial, to stop and ask when requirements conflict, to avoid unrequested changes, and to back up claims of completion with evidence such as passing tests or build output.

• Routes a task to a skill by development phase, for example spec-driven-development for a new feature or debugging-and-error-recovery when something broke.
• Asks the agent to surface assumptions before coding and to name confusion instead of guessing.
• Limits scope to what was asked and favours simple solutions.
• Requires verification before work is called done.

Install the skill from the ZIP download

Download the ZIP for the skill from this page and unzip it. The skill is a folder that contains a SKILL.md file. Place the folder in one of two locations depending on how you want to use it.

For a personal install available to all projects, unzip into your user skills folder at ~/.claude/skills/. For a single project, unzip into that project's .claude/skills/ folder so it applies only there.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.

• Personal (all projects): ~/.claude/skills/
• Single project: .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need a special command. The skill's description says it is used when starting a session or when you need to decide which skill or workflow applies. In practice, you describe the work in plain language and the agent decides whether to consult this routing guide.

If you want to be explicit, name the phase or the outcome you need. Naming the skill is the most direct way to be sure it is used, but the wording of the description suggests the agent will also reach for it at the start of a session on its own.

• "Before we start, figure out which of our skills applies to adding a login page."
• "This feature is non-trivial. Use the using-agent-skills routing and tell me what you'd do first."
• "Something broke in the export job. Which skill should we use to track it down?"
• "We're ready to ship this. What's the right skill sequence from here?"

Examples of what you will see

Because the skill sets a habit of stating assumptions, a typical response to a non-trivial request starts with a list of assumptions and a request for correction, rather than immediate code. The skill's own template for this looks like the following.

When the agent finds conflicting requirements, it is expected to name the specific conflict and wait for an answer rather than picking one silently.

• For a bug fix, the routing points to debugging-and-error-recovery, then test-driven-development, then code-review-and-quality.
• For a full feature, the routing lists a sequence that starts with interview-me or spec-driven-development and ends with shipping-and-launch.
• For a security concern during review, the routing points to security-and-hardening.

What the skill needs and what it does not do

The skill requires an agent that reads SKILL.md files, such as Claude Code or a compatible runtime. For the routing to have any effect, the other skills in the Agent Skills collection must also be installed, because this skill only points to them.

The skill does not run scripts and does not use the network. It is a set of instructions, so it does not perform a task itself. It also refers to a definition-of-done document, which is not included in this download, so that reference will not resolve from the skill alone.

• Requires: an agent that reads SKILL.md files, plus the other collection skills for routing to work.
• Runs scripts: no.
• Uses network: no.
• Files: SKILL.md.

Safety notes

Because the skill runs no scripts and makes no network calls, it has no direct side effects on your system. Its influence comes from the instructions it gives the agent, so the agent's behaviour is still what you should review.

Be aware that the skill tells the agent to avoid deleting code it does not understand and to avoid unrequested refactoring. Review changes the agent proposes, particularly on the skills it routes to that may act on code or deployments.

• No scripts run and no network access is used by this skill itself.
• The referenced skills are separate installs and should be reviewed on their own terms.
• Check the agent's changes before accepting them, as you would with any agent output.

Tips for getting the most from it

Install the other skills in the collection first. Without them, the routing has nothing to point to. Start with the skills for the phases you use most.

For small jobs, you can skip the process. The skill itself notes that simple one-line edits may not need the full overhead, so ask for a quick change directly.

• Use it at the start of a session to set the working process.
• Answer the assumptions list quickly; corrections are cheaper before coding than after.
• Not every task needs every skill. A bug fix may need only three.