fixlisted
Install: claude install-skill Smart-AI-Memory/attune-ai
# Fix (guided intake)
**IMPORTANT: Start your response by telling the user:**
> **Fix** — Composing an outcome-first fix: goal, scope, and
> verification probes, then a preview before anything runs.
## What It Does
Interactive intake for `attune fix` (the outcome-first Fix surface,
`docs/specs/outcome-first-fix/`): one form gathers the goal, the
`--scope` the diff must stay confined to, and the `--probe`
commands that verify the fix — with scope and probe options DERIVED
from the working tree (changed paths and matching test files), not
typed from memory. The composed CLI command is previewed before any
execution; the receipt independently verifies every probe.
Relationship to `/fix-test`: that skill diagnoses and fixes a
FAILING TEST in-session. This skill drives the `attune fix` CLI
contract — goal + scope + probes + receipt — for any code fix.
Neither replaces the other.
## Step 1 — Derive candidates and build the form
```bash
python -m attune.elicitation.fix_intake
```
The JSON payload contains a validated form definition
(`attune.elicitation.fix_intake.build_fix_intake_form`) plus the
derived `scopes` and `probes` lists. Changed paths lead; on a clean
tree the candidates fall back to recently-touched directories from
git history, so pickers render in either state — empty lists mean
the repo has no usable history at all. If the user's invocation
already stated the goal, carry it into the request field as the
default rather than asking again.
## Step 2 — Render th