skill-bench
Skill0 starsChanged 29 days agoMITView on GitHub
Runs a skill's evals in every installed AI tool, side by side. Use for "does it work in Codex or Gemini too", or before publishing to people on other tools.
- Installs packages
3 other projects use a version of this.
Adopted by
Most recently taken up by
- busycaesar/Harnesstheir own version, first seen Nov 1, 2026
- tammai/bigin-skillstheir own version, first seen Oct 11, 2026
- thoitiettxl-cyber/Claw-IItheir own version, first seen Apr 3, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- thoitiettxl-cyber/Claw-II★ 2same name
- busycaesar/Harness★ 1same name
- tammai/bigin-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skill-checkSkill
- skill-competeSkill
- skill-createSkill
- skill-descriptionSkill
- skill-evalSkill
- skill-factsSkill
- skill-improveSkill
- skill-learnSkill
- skill-packageSkill
- skill-planSkill
- skill-pluginSkill
- skill-publishSkill
- skill-releaseSkill
- skill-repoSkill
- skillSkill
- skill-tidySkill
- skill-tokensSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Run check first: anything it flags as one tool's feature will show up here as a gap.
- Distinctive: line 19
Read the table with them: for each tool, the share passed with and without the skill, the difference, the time, and how often the skill …
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The same tests in every tool
A skill written in one tool often half-works in another: a path one tool doesn't have, a feature it ignores, a description its list cuts. This runs the skill's own evals in each installed tool so "it works in Codex too" is a number.
The engine runs as npx -y @thenavidm/skill-builder@latest. Where skill-builder is installed (npm install -g @thenavidm/skill-builder), run skill-builder instead: it's the same engine, without the download.
- Run
checkfirst: anything it flags as one tool's feature will show up here as a gap. - Say how many runs it takes (tasks x 2 x runs x tools), then:
npx -y @thenavidm/skill-builder@latest bench <skill> --runs 3- Read the table with them: for each tool, the share passed with and without the skill, the difference, the time, and how often the skill never ran. Then the tasks where the tools disagree.
When a tool falls behind
| Pattern | Look at |
| The skill never ran in one tool | Its description in that tool's crowded list: tokens, then skill-description |
| It ran but failed in one tool | A path or feature only another tool has: check's tool-only findings |
| Runs were blocked | That tool asked for a permission headless mode can't give; the result says which runs |
| Gemini CLI didn't run | It refuses free personal accounts (October 2026); it needs a paid or API-key account |
The skill-analyzer agent can read all the tools' results together and name the likely cause of each gap. Each tool has its own review page under the output folder. Fix in the skill, then run the bench again on the tool that fell behind.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.