skill-improve
Skill7 starsChanged 48 days agoMITView on GitHub
Reads a skill's SKILL.md plus its open friction-log entries and proposes concrete diffs to make it more effective and efficient. Read-only until approval: applies only diff by diff, never rewrites wholesale, and never creates a new skill from scratch (that is /skill-detector). Use when the operator says: /skill-improve, skill-improve, improve the X skill, evolve the X skill.
15 other projects use a version of this.
Adopted by
Most recently taken up by
- thenavidm/skill-buildertheir own version, first seen Nov 8, 2026
- 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
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
- hongyuanc/codex-game-studios★ 5same name
- gonimar/claude-web-studio★ 4same name
- hideki5123/agent-skill-set★ 2same name
- SebastianLuser/claude-code-arcane★ 2same name
- dawn840705/claude-code-studios★ 1same 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 repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- daily-log/CLAUDE.mdCLAUDE.md
- memory/CLAUDE.mdCLAUDE.md
- operators/CLAUDE.mdCLAUDE.md
- daily-briefingSkill
- handoverSkill
- html-previewSkill
- meeting-debriefSkill
- memory-checkupSkill
- pr-reviewSkill
- session-debriefSkill
- setup-company-brainSkill
- shipSkill
- skill-detectorSkill
- sparringSkill
- syncSkill
- system-checkupSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 13 edited versions we read keep these.
2
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 47
Read skills/_improvements/friction-log.md and keep only the open entries for that skill: entries above the ## Archive marker whose header …
- Distinctive: line 82
No open entries for that skill: say so and stop.
- Very specific: line 110
If the team routes structural work through feature branches (see docs/GOVERNANCE.md step 2 if the team has adopted it), skill edits belong …
- Distinctive: line 117
Additive first: Do not touch logic that already works unless it is the recorded cause of the friction.
“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
Skill Improve
A meta-skill: it turns a skill's recorded friction into an approvable improvement proposal. It never self-applies. It proposes a diff, the operator decides, diff by diff.
Input
/skill-improve <skill-name>, e.g. /skill-improve sync. If the name is
omitted, rank skills by number of open friction-log entries and ask which
one to take. The count reads the open section only (above ## Archive)
and flags entries whose skill name matches no folder under skills/;
misfiled open entries are surfaced by the Step 1 archive check, not ranked
here:
cd "$(git rev-parse --show-toplevel)"
awk '/^## Archive/{exit} /^## [0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9] /{print}' skills/_improvements/friction-log.md \
| sed -E 's/^## [0-9-]+ · ([^ ]+) · .*$/\1/' | sort | uniq -c | sort -rn \
| while IFS= read -r line; do
name="$(printf '%s\n' "$line" | sed -E 's/^ *[0-9]+ //')"
if [ -d "skills/$name" ]; then echo "$line"; else echo "$line OWNERLESS (no skills/$name/)"; fi
doneAn OWNERLESS line is not a skill to improve; it is an entry to route
(Step 1, "Ownerless entries"). Present those separately from the ranking.
Workflow
1. Gather
- Find and read the target skill's
SKILL.mdatskills/<skill-name>/SKILL.md(the suite is flat: one directory per skill, no category subfolders), plus any supporting file it references, if relevant to the friction. Read
skills/_improvements/friction-log.mdand keep only the open entries for that skill: entries above the## Archivemarker whose header names the skill. Open is defined by the entry, not by its position: an entry with noconsumed:line is open wherever it sits. Check the archive for misfiled ones, and treat any it finds as open, saying so to the operator:H="$(awk '/^## Archive/{a=1} a' skills/_improvements/friction-log.md | grep -cE '^## [0-9]{4}-[0-9]{2}-[0-9]{2} ')" C="$(awk '/^## Archive/{a=1} a' skills/_improvements/friction-log.md | grep -c '^consumed:')" [ "$H" = "$C" ] && echo "archive ok: $H entries, all consumed" || echo "MISFILED: $H archived entries, $C consumed lines; the difference is open"- Ownerless entries. An entry whose skill name matches no folder under
skills/(a cross-skill pattern, an external plugin, the harness) is reached by no/skill-improve <name>run. When you meet one, route it to exactly one of three places and say which: (a) a pattern that recurs across skills becomes a class inskills/_improvements/known-patterns.md, plus one line citing that class from an executable step of every skill it touches, because a note no step opens closes no loop; (b) an entry whose tag names another skill of the suite is refiled under that skill's name; (c) a cause outside the repo is archived as not fixable in-repo, with the substitute habit written in itsconsumed:line. The rules are indocs/SKILL-MAINTENANCE.mdsection 5.
2. Diagnose
- Group the open entries by pattern: same step, same recurring trigger.
Then open
skills/_improvements/known-patterns.mdand check every group against its classes; an entry that already carries apattern:line names its class. A group that matches a class is a suite-wide pattern, not a defect of this one skill: the proposal names every skill the class touches, and the fix is usually the same line in each. - A recurring pattern outranks a one-off; say so explicitly and rank patterns above isolated entries.
- No open entries for that skill: say so and stop. No proposal without data. The only exception is a light qualitative review, and only if the operator explicitly asks for one.
3. Propose (on screen, write nothing yet)
For each pattern, present:
- Symptom: what happens and how often, citing the entry dates.
- Cause: which part of the SKILL.md causes it.
- Proposed diff: the exact change to the SKILL.md, before to after, aimed at both effectiveness and efficiency.
- Cost/risk: what could break. Conservative default: additive change before rewrite.
Present every proposal, then stop and wait for the operator's approval, diff by diff.
4. Apply approved diffs only
- For each accepted proposal, edit the SKILL.md with a targeted edit, never a full rewrite of the file.
- Close each consumed entry by appending one line to it,
consumed: YYYY-MM-DD · <what changed, one line>, then move the whole entry from the open section to below the## Archivemarker. The line is the marker of a closed entry; the move is housekeeping. Never move an entry without the line, and never add the line to an entry whose change was not applied. - If the team routes structural work through feature branches (see
docs/GOVERNANCE.mdstep 2 if the team has adopted it), skill edits belong on one, not on a direct commit to main.
Rules
- Never apply without explicit approval, judged diff by diff.
- Additive first. Do not touch logic that already works unless it is the recorded cause of the friction. The goal is to refine, not rewrite.
- No proposals without data. An empty log for that skill means no diff.
- One skill at a time.
- Never modify a skill's
## Self-improvementfooter. It is standard and identical across the suite; changing it is a change toskills/_improvements/capture.md, a different decision than this one.
Self-improvement
At the end of the run, check it against these friction triggers: avoidable
round-trip · improvisation · breakage · token waste · operator correction
(full definitions in skills/_improvements/capture.md). If at least one fired,
ask the operator "there was friction on X, should I log it?"; on their ok,
append an entry to skills/_improvements/friction-log.md in the format
capture.md defines. Clean run: say nothing. Never self-edit this skill;
improvements go through /skill-improve.
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.

