git:20260721.fe3067c to git:20260721.9534dd9

2 added, 2 removed. Audit A to A.

# AI-SKILLS

This repository is a library of AI-agent **skills** (working methods to follow)

and **prompts** (ready-to-run templates). It is meant to be operated

autonomously: you do not wait for the user to name a skill.

For any task, run the routine below on your own:

1. Name the task's intent in a phrase (review code, write an email, debug an

error, design a system, research a question).

2. Open [`index.json`](index.json) and match your intent against each skill's

`use_when` trigger or each prompt's `description`.

3. Take the 1-3 entries that genuinely fit; a strong single match beats three

loose ones.

4. Load a skill's `SKILL.md` and follow its method, or fill a prompt's

`{variables}` and run it.

- 5. Check your output against the entry's own `## Boundaries` or rules before

- finishing.

+ 5. Check your output against the entry's own guardrail section (`## Boundaries`,

+ or an equivalent like `## Rules` or `## Litmus tests`) before finishing.

The full protocol, including how to compose entries across a task's phases,

is below.

@AGENTS.md