---
name: gavel-gain
description: >
  Suite-health scoreboard from test run results: pass rate, flake count, LOC per
  test, coverage delta. Report-only — does not provision environments or track
  issues. Trigger: /gavel-gain, "test quality scoreboard".
---

# Gavel Gain

Display this scoreboard when invoked. One-shot: do NOT change mode, write flag
files, or persist anything.

The figures come from the current repo's test run results, not benchmarks.
Parse the latest test report (Playwright JSON, pytest JUnit XML, etc.) to
compute real metrics.

## Scoreboard

Render plain ASCII:

```
  gavel gain                        test quality scoreboard

  Tests           total: <N>  passed: <P>  failed: <F>  skipped: <S>
  Pass rate       ████████████████████  <P/N * 100>%
  Flake count     <flaky tests identified in last N runs>
  LOC per test    <avg lines per test file>
  Constitution    <violations found by last /gavel-audit>
  Coverage delta  <new tests added vs removed since last run>

  This repo:  /gavel-debt   (deferrals you marked)
              /gavel-audit  (what's still cuttable)
              /gavel-review (diff-level violations)
```

## Per-Framework Parsing

- **Playwright**: parse `test-results/*.json` or `--reporter=json` output       
- **pytest**: parse JUnit XML from `--junitxml` or `pytest-json-report`
- **JUnit/TestNG**: parse Surefire/Failsafe XML reports
- **Cypress**: parse `cypress/results/*.json` from `cypress-multi-reporter`     
- **WebdriverIO**: parse `wdio` reporter output

**Flake count** comes from `gavel flakiness <report>` — it scores retry counts
and pass/fail flips per test (`flakyCount`). Report-only; a test that fails every
retry is a real failure, not flake.

## Honesty boundary

These are real numbers from the current repo, not benchmarks. If no test
results are found, say so: `No test results found. Run your suite first.`

NEVER fabricate metrics. If the data isn't there, say so.

## Boundaries

One-shot display. Edits nothing, changes no mode.
"stop gavel" or "normal mode": revert.