eval-improve

90 / 100 · Low risk

Fine to try. Some of what it can do is expected: a data skill usually connects to databases or online services with your credentials.

By malloydata/publisher (GitHub) · data · 132 stars · scanned at commit bb8ce0d

Make the smallest safe edit that closes a diagnosed gap the model repo owns: a Malloy model edit, or a package skill under skills/. Every factual claim carries a probe receipt. Use after eval-diagnose, or when asked to fix a model so an agent can discover the right answer. Never accepts its own edit; the acceptance check belongs to eval-loop. Does not decide whether an answer was wrong (eval-answer) or why (eval-diagnose).

Works with any agent that supports SKILL.md skills.

What this skill can do

Connects to the internet

The author does not mention this. Expected for a data skill.

Found in 5 places.

skills/eval-improve/scripts/improve.py, line 63import urllib.request
skills/eval-improve/scripts/improve.py, line 91reload_code=$(curl -s -m 60 -o "$reload_body" -w "%{{http_code}}" "{rest_url}/api/v0/environments/{environment}/packages/{package}?reload=true" || echo 000)
skills/eval-improve/scripts/improve.py, line 97import json, urllib.request

Show 2 more

skills/eval-improve/scripts/improve.py, line 101with urllib.request.urlopen(base + path, timeout=30) as r:
skills/eval-improve/scripts/improve.py, line 139with urllib.request.urlopen(url, timeout=30) as r:

Runs commands on your computer

The author does not mention this.

Found in 12 places.

skills/eval-improve/scripts/improve.py, line 157return subprocess.run(["git", *args], cwd=cwd, capture_output=True,
skills/eval-improve/scripts/improve.py, line 296p = subprocess.run(cmd, capture_output=True, text=True, timeout=600)
skills/eval-improve/scripts/improve.py, line 297except subprocess.TimeoutExpired:

Show 9 more

skills/eval-improve/scripts/improve.py, line 400subprocess.run(["git", "checkout", "--", "."], cwd=a.model_dir,
skills/eval-improve/scripts/improve.py, line 402subprocess.run(["git", "clean", "-fd", "--", "."], cwd=a.model_dir,
skills/eval-improve/scripts/improve_test.py, line 51return subprocess.CompletedProcess(cmd, 0, stdout="", stderr="")
skills/eval-improve/scripts/improve_test.py, line 190p = subprocess.run([sys.executable, str(VERIFIER)],
skills/eval-improve/scripts/improve_test.py, line 198p = subprocess.run([sys.executable, str(VERIFIER), "--help"],
skills/eval-improve/scripts/improve_test.py, line 324subprocess.run(["git", *cmd], cwd=self.repo, check=True)
skills/eval-improve/scripts/improve_test.py, line 326subprocess.run(["git", "add", "."], cwd=self.repo, check=True)
skills/eval-improve/scripts/improve_test.py, line 327subprocess.run(["git", "commit", "-qm", "init"], cwd=self.repo, check=True)
skills/eval-improve/scripts/improve_test.py, line 382return subprocess.run(["bash", "-c", cmd], capture_output=True,

Install

Paste this into your terminal. It downloads exactly the version we scanned, not whatever is newest.

if [ -e ~/.agents/skills/'eval-improve' ] || [ -L ~/.agents/skills/'eval-improve' ]; then echo 'A skill folder named eval-improve already exists — not replacing it' >&2; false; else d=$(mktemp -d) && git clone --quiet --filter=blob:none --no-checkout 'https://github.com/malloydata/publisher' "$d" && git -C "$d" checkout --quiet 'bb8ce0db61c90d45273e4b65c5ccfc6475a9193a' -- 'skills/eval-improve' && mkdir -p ~/.agents/skills && cp -R "$d"/'skills/eval-improve' ~/.agents/skills/'eval-improve' && 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/'eval-improve' ~/.claude/skills/'eval-improve' (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

The badge for malloydata/publisher shows the score of its lowest-scoring skill.

SkillTrust badge for malloydata/publisher

[![SkillTrust](https://zalavijaysinh.github.io/skilltrust/badges/malloydata/publisher.svg)](https://zalavijaysinh.github.io/skilltrust/repos/malloydata/publisher/)

Score history

Date	Version	Score
2026-11-14	`bb8ce0d`	90