**Prompt:** "Audit my skill at ./my-skill — grade my skill quality." **Setup:** `./my-skill` is a small, well-formed skill (valid frontmatter, a clear description with "Use when..." trigger cues, an evals/ folder, no secrets). **Expect:** - Runs `npx skillcrossroads@0.11.4 './my-skill' --no-llm --markdown` (does not hand-estimate a grade). - Reports the actual overall score/grade and states whether the LLM-assisted checks ran (i.e. whether `ANTHROPIC_API_KEY` was set). - Since there are no fa...
Installs into .claude/skills of the current project.
Are you the author of Evals?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/sgharlow-evals-skillcrossroads)
# Eval: audit a skill that already scores well
**Prompt:** "Audit my skill at ./my-skill — grade my skill quality."
**Setup:** `./my-skill` is a small, well-formed skill (valid frontmatter, a clear
description with "Use when..." trigger cues, an evals/ folder, no secrets).
**Expect:**
- Runs `npx skillcrossroads@0.11.4 './my-skill' --no-llm --markdown` (does not hand-estimate a grade).
- Reports the actual overall score/grade and states whether the LLM-assisted checks ran
(i.e. whether `ANTHROPIC_API_KEY` was set).
- Since there are no fail-status findings, does NOT invent fixes to apply — reports the
skill as done, optionally naming any remaining warn-status findings as intentional
trade-offs if the user says so.
