# Skill Autobench

> | Author an eval for an existing skill from its REAL usage history, not its spec. Mine invocations from the brain's conversation archive (conversations/) and per-harness session transcripts — a user correction after an invocation is the gold signal — then synthesize an evalcontract plus 4-8 replayable cases with honesty labels (SPEC-DERIVED vs HISTORY-IMPLIED) and stage the result at skills/<name/eval/autobench-<date.md as PENDING-HUMAN-APPROVAL. Never rewrites SKILL.md. Ships two guard companions: panel integrity (multi-model judging must prove each provider actually responded) and the fail-improve taxonomy (logged LLM-fallback cases convert to deterministic code over time).

## Facts
- Page: https://tashan.sh/capability/skill-cheroma-skill-autobench
- tashan id: skill:cheRoma/skill-autobench
- Source: https://github.com/cheRoma/gbrain-link-chat
- Type: skill
- Category: devtools
- tashan score: 50.0 / 100
- Adoption: 14.0
- Upkeep: 98.0
- Freshness: 96.0
- Evidence coverage: 84% of the inputs this score can use
- Health: active
- Instruction depth: not yet graded
- License: MIT
- Official: no

## Install

```sh
cp -r skill-autobench ~/.claude/skills/
```

## Security audit
Not scanned. We audit npm-published capabilities; this one has no npm package we can resolve, or has not reached the queue. This is not a clean bill of health.

---
Measured 2026-09-12 by tashan (https://tashan.sh) from public evidence. Scorer s5.