done-check-and-eval
100 / 100 · Low risk
Fine to try.
What the expert builds first, with the runnable check the scaffold uses to decide done kept apart from the supplied evaluator, the specification the scaffold gives the model, fixed conditions against the settings the scaffold may change, pass^1 and pass^k, the replication noise between two identical runs, paired comparisons and task-level flips, the sealed held-out rule, the fast subset, episode isolation, failure attribution, the results convention results/episodes.jsonl written by an adapter …
What this skill can do
We found nothing in the files we read that matches our rules. That does not prove the skill is safe: see the limits of static scanning.
Install
Paste this into your terminal. It downloads exactly the version we scanned, not whatever is newest.
if [ -e ~/.agents/skills/'done-check-and-eval' ] || [ -L ~/.agents/skills/'done-check-and-eval' ]; then echo 'A skill folder named done-check-and-eval already exists — not replacing it' >&2; false; else d=$(mktemp -d) && git clone --quiet --filter=blob:none --no-checkout 'https://github.com/Leeroo-AI/kapso' "$d" && git -C "$d" checkout --quiet '35f7b9b26a6b3864fdc8c018577d448c2840345b' -- 'src/kapso/harness/bundled/agent-engineering/skills/done-check-and-eval' && mkdir -p ~/.agents/skills && cp -R "$d"/'src/kapso/harness/bundled/agent-engineering/skills/done-check-and-eval' ~/.agents/skills/'done-check-and-eval' && rm -rf "$d"; fi
Codex reads ~/.agents/skills directly. For Claude Code, Cursor, Copilot, Gemini CLI, Antigravity or OpenCode, also link it. For Claude Code: ln -s ~/.agents/skills/'done-check-and-eval' ~/.claude/skills/'done-check-and-eval' (other agents: ~/.cursor/skills, ~/.copilot/skills, ~/.gemini/skills, ~/.gemini/config/skills, ~/.config/opencode/skills). Or install through the finder skill, which does both.
Badge for your README
[](https://zalavijaysinh.github.io/skilltrust/repos/Leeroo-AI/kapso/)
Score history
| Date | Version | Score |
|---|---|---|
| 2026-11-13 | 35f7b9b | 100 |