recoup-internal-eval-skill-benchmark
Installation
SKILL.md
Skill Benchmark
Grade any plugin or skills folder against the frontier bar for what makes skills and plugins good. The output is an evidence-backed report: where the pack sits, what's fat vs. thin, what should be combined/split/parameterized, and the prioritized, boundary-carrying moves worth stealing.
This skill follows the very discipline it grades: a deterministic measurement
script does the counting (so numbers are never eyeballed), and the latent
judgment — scoring, diagnosis, recommendations — is the agent's job, grounded in
the reference standard. The script and references ship alongside this SKILL.md;
invoke them with the relative paths below.