Example audit

One paragraph: what problem this solves and what the output is.

When to Use

• Use for the authorized assessment described above.
• Use when the named artifact and scope are available.

When NOT to Use

• Do not treat the skill as permission to test or change a target.
• Do not turn missing evidence into a clean result.

Workflow

1. Numbered steps. Each one is an action with a visible result.
2. Add a script only when it performs real reusable work.
3. Move conditional detail to references/ when progressive disclosure makes the core workflow easier to follow.

Judgment calls

Where the model has to decide something, say what the decision depends on and what the trade-off is. Say when not to do the thing.

Rationalizations to reject

Required for security categories; useful everywhere. Each entry: the shortcut, in the voice someone would use to justify it, then why it is wrong.

<!-- Security skills must retain this section. See AGENTS.md. -->