skill-improve
Skill0 starsChanged 29 days agoMITView on GitHub
Fixes a skill that gets something wrong, proven on the failing case and on everything it did right before. Not for tidying (skill-tidy) or new skills (skill-create).
- Installs packages
15 other projects use a version of this.
Adopted by
Most recently taken up by
- striderZA/OpenCodeGameStudiostheir own version, first seen Nov 4, 2026
- Donchitos/Claude-Code-Game-Studiostheir own version, first seen Oct 19, 2026
- venus2099/codex-game-studiostheir own version, first seen Oct 17, 2026
- YOSSII812001/claude-code-jarvistheir own version, first seen Oct 9, 2026
- HunLuanZhiZhu/Game-Studios-Plugintheir own version, first seen Oct 5, 2026
- marcogalluccio/company-brain-ostheir own version, first seen Sep 28, 2026
Used with
Projects with it also use soak-test, team-live-ops and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
lilangMaxSkillsoak-test
Generate a soak test protocol for extended play sessions. Defines what to observe, measure, and log during long play sessions to surface slow leaks, fatigue effects, and edge cases that only appear after sustained play. Primarily used in Polish and Release phases.
Also in 9 projects that use this
lilangMaxSkillteam-live-ops
Orchestrate the live-ops team for post-launch content planning: coordinates live-ops-designer, economy-designer, analytics-engineer, community-manager, writer, and narrative-director to design and plan a season, event, or live content update.
Also in 9 projects that use this
- +10 more from lilangMax
Other versions15 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 15
- Donchitos/Claude-Code-Game-Studios★ 27ksame name
- striderZA/OpenCodeGameStudios★ 118same name
- cenconq25/claude-code-app-studio★ 40same name
- dawn840705/codex-code-studios★ 17same name
- lilangMax/ClaudeCodeGameStudios★ 8same name
- marcogalluccio/company-brain-os★ 7same name
- hongyuanc/codex-game-studios★ 5same name
- gonimar/claude-web-studio★ 4same name
- hideki5123/agent-skill-set★ 2same name
- SebastianLuser/claude-code-arcane★ 2same name
5 more not listed
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-benchSkill
- skill-checkSkill
- skill-competeSkill
- skill-createSkill
- skill-descriptionSkill
- skill-evalSkill
- skill-factsSkill
- 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
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 13 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
The exact failure: Ask for the real input and the output that was wrong, or their review notes (feedback.json from an eval run).
- Distinctive: line 19
The cause: Read the whole skill and find the instruction that produced it, the rule that's missing, or the description that let another …
- Very specific: line 21
Prove it: Add the failing case to evals/evals.json as a task with an assertion that failed before.
“Common” is measured: kept by at least 6 in 10 of the 13 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
Improve a skill
One standard: prove it. An edit that only looks right, with the hope that it works next time, hasn't fixed anything. To clean up a skill that works, use skill-tidy.
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.
First, save the current version
Copy the whole skill folder to .agents/skill/history/<name>/<date-and-time>/ and confirm the copy matches. It is the baseline the new version is measured against, and the way back if the change makes things worse. If the skill came from someone else's plugin, ask before changing it: edit a copy under a name of theirs, or suggest the fix to the plugin's author.
Fix it
- The exact failure. Ask for the real input and the output that was wrong, or their review notes (
feedback.jsonfrom an eval run). "It's off" isn't enough to fix; "it wrote the date as November 10" is. - The cause. Read the whole skill and find the instruction that produced it, the rule that's missing, or the description that let another skill fire. Say it in one sentence, pointing at the line. A guess isn't a cause.
- The smallest change that removes the cause, written by the
skill-writeragent from your one-sentence cause, or by you in a tool without agents. A skill that grows a new rule for every case becomes a list nobody can follow. Prefer explaining the reason behind an existing rule over adding a new one. - Prove it. Add the failing case to
evals/evals.jsonas a task with an assertion that failed before. Runevalon the new version and on the saved one:
npx -y @thenavidm/skill-builder@latest eval <skill> --runs 3The failing task must now pass, and every task that passed before must still pass. Where assertions can't capture the difference, give the old and new outputs to the skill-comparator agent unlabelled; then have the skill-reviewer agent check the changed skill. If the cause was triggering, use skill-description instead and prove it with trigger.
Either way
Two attempts. If the second still fails its proof, put the saved version back (copy the whole folder, confirm it matches) and say plainly what's left: the fix may need a rebuild with skill-create. Report what changed, the before and after numbers, and where the old version is kept.
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.

