git:20260808.623abbc to git:20260809.d5365e0

# research-evaluation skill

Evaluates external tools/projects before adoption: identifies each target (web

search, NOT_FOUND allowed), runs the seven-dimension rubric (value · setup/licensing

· adversarial · security · stability · quality · docker-or-skip demo), and lands the

results as research design docs + tracking issues in the repo's `docs/mbo/` system

(or a `docs/research/` fallback), fanned out to parallel agents for multi-target

batches.

headroom, claude-code-setup, task-observer). `evals/evals.json` holds trigger cases.