What do I need to use Agent Spec?

Job to be done; Tools / actions; Autonomy level; Risk surface; Success definition & escalation.

Is Agent Spec free to use?

Yes — run it free in the browser playground (no API key needed for the sponsored trial, or bring any provider key), or install it into Claude Code, Cursor, ChatGPT and more via npx pm-claude-skills. MIT-licensed and open source.

Agent Spec

Q: How do I know the Agent Spec output is good?

The skill verifies itself against its own quality checks: Every tool is marked reversible/irreversible, and every irreversible action has a human gate; There is a hard max-steps and max-cost budget — the loop cannot run unbounded; Escalation triggers are explicit (confidence, repeated failure, out-of-scope, high-risk); The safe default on uncertainty is "stop and ask", not "guess and act".

🔵 Stable🕐 updated 2026-06-24 🔷 SkillSpec L3 pm-ai

Specify an autonomous or tool-using AI agent before building it. Use when asked to design an AI agent, define an agent's tools and guardrails, scope what an agent is allowed to do, or write an agent spec/PRD. Produces an agent spec — goal & scope, tools with permissions, the control loop, guardrails & approval gates, memory, escalation/handoff, evaluation, and failure handling.

▶ Run it free — no key needed 📝 Grade your existing draft View SKILL.md ↗

What to give it

▸Job to be done — the outcome the agent owns, and the boundary of its authority.

▸Tools / actions — what it can call (read APIs, write actions, code execution), and which are irreversible.

▸Autonomy level — fully autonomous, propose-then-approve, or co-pilot.

▸Risk surface — what's the worst thing a wrong action could do (spend money, send a message, delete data)?

▸Success definition & escalation — how "done" is judged, and when it must hand off to a human.

✅ The bar it holds itself to

Every skill in this library self-verifies — these are this skill's own quality checks, straight from its definition.

✓Every tool is marked reversible/irreversible, and every irreversible action has a human gate

✓There is a hard max-steps and max-cost budget — the loop cannot run unbounded

✓Escalation triggers are explicit (confidence, repeated failure, out-of-scope, high-risk)

✓The safe default on uncertainty is "stop and ask", not "guess and act"

✓Evaluation includes a safety metric (wrong/unauthorised actions), not just task success

✓Non-goals and authority limits are stated, not implied

⚠️ What it refuses to do

Do not give an agent irreversible actions without an approval gate — autonomy and irreversibility together is how agents cause real damage

Do not omit a step/cost budget — an agent that can loop is an agent that can rack up cost or thrash forever

Do not measure only task success — an agent that completes the task by taking a wrong action has failed

Do not let the agent invent tool calls or arguments — validate against the schema and fail safe

Do not skip the "what's the worst case" analysis — the risk surface determines how many guardrails you need

Install

npx pm-claude-skills add --agent claude   # or codex · cursor · gemini · hermes
# or one-line MCP (every skill, any client):
claude mcp add pm-skills -- npx -y pm-claude-skills-mcp

Related skills

🔌 Embed this skill

Drop this on your blog, docs, or site — it renders a "Run this skill" card:

<div data-pm-skill="agent-spec"></div>
<script src="https://mohitagw15856.github.io/pm-claude-skills/embed.js" async></script>

💬 Discussion

Agent Spec is one of 551 open-source professional AI agent skills — all SkillSpec L3. Try them all in the browser · ⭐ Star on GitHub · Browse the full catalog