agent-spec

1# Agent Spec Skill
2 
3An agent is a model plus tools plus a loop — and the danger lives in the tools and the loop, not the
4model. This skill specifies an agent so its *authority is explicit*: what it can do, what needs a human
5yes, and what happens when it's wrong. Scope and guardrails first; cleverness second.
6 
7> **Not quite this?** Use `agent-observability-spec` when the agent is in production and needs tracing and alerting.
8 
9## Required Inputs
10 
11Ask for these only if they aren't already provided:
12 
13- **Job to be done** — the outcome the agent owns, and the boundary of its authority.
14- **Tools/actions** — what it can call (read APIs, write actions, code execution), and which are irreversible.
15- **Autonomy level** — fully autonomous, propose-then-approve, or co-pilot.
16- **Risk surface** — what's the worst thing a wrong action could do (spend money, send a message, delete data)?
17- **Success definition & escalation** — how "done" is judged, and when it must hand off to a human.
18 
19## Output Format
20 
21### Agent Spec: [name]
22 
23**1. Goal & scope** — the job in one sentence; explicit non-goals and authority limits.
24 
25**2. Tools / actions** — a table; mark each action's reversibility and required permission.
26 
27| Tool | Purpose | Reversible? | Gate |
28|---|---|---|---|
29| search_kb | read context | yes | none |
30| send_email | notify | **no** | **human approval** |
31…