skill-profiles
Skill2 starsChanged 3 days agoMITView on GitHub
Turn groups of skills on or off depending on the type of work (frontend, marketing, video, code, ...) so the skill listing stays within the context budget. Use when the person says what kind of work they'll do ("today is frontend", "let's do video"), asks to see/change/create profiles, or asks why a skill doesn't show up or why the skills listing is truncated. Claude Code only.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- anti-slopSkill
- handoffSkill
- weekly-updateSkill
- youtube-scriptwritingSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 35
Read skill-profiles.json and resolve the active set (union of the profile's groups).
- Very specific: line 37
Update active in skill-profiles.json.
- Distinctive: line 40
Always merge the full JSON with a script or temp file.
- Very specific: line 44
Only skills in ~/.claude/skills/.
- Very specific: line 45
Never delete files: a profile only writes skillOverrides.
- Distinctive: line 46
If the work doesn't fit a profile, propose the closest one and offer to create a new one.
- Very specific: line 47
If they ask to enable a single skill that's off, remove its entry from skillOverrides without changing the rest.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill profiles
Claude Code loads every skill's description at startup, with a small budget (~1% of the context). With many skills, descriptions get truncated and the agent stops picking them by itself. A profile leaves active only the skills for the current work.
Catalog
File ~/.claude/skill-profiles.json (if it doesn't exist, create it with the structure below and ask the person which groups they want based on the skills in ~/.claude/skills/).
{
"active": "all",
"groups": {
"core": ["handoff", "perfiles", "find-skills", "last30days", "humanizer"],
"frontend": ["design-taste-frontend", "high-end-visual-design", "emil-design-eng", "apple-design"],
"marketing": ["copywriting", "content-strategy", "social", "cold-email", "ads", "seo-audit"],
"video": ["hyperframes", "hyperframes-cli", "media-use"]
},
"profiles": {
"code": { "desc": "Essentials only", "groups": ["core"] },
"frontend": { "desc": "Interface design", "groups": ["core", "frontend"] },
"marketing": { "desc": "Marketing and content", "groups": ["core", "marketing"] },
"video": { "desc": "Video production", "groups": ["core", "video"] },
"all": { "desc": "Everything on", "groups": ["core", "frontend", "marketing", "video"] }
}
}Apply a profile
- Read
skill-profiles.jsonand resolve the active set (union of the profile's groups). - In
~/.claude/settings.json, writeskillOverrideswith"off"for every skill in~/.claude/skills/that isn't in the active set, and no entry for the active ones. Don't touch other keys. Back up first. - Update
activeinskill-profiles.json. - Tell the person: the change shows after restarting the session.
Always merge the full JSON with a script or temp file; never interpolate skill names inside a shell command.
Rules
- Only skills in
~/.claude/skills/. Don't touch plugin or built-in ones. - Never delete files: a profile only writes
skillOverrides. Going back toallturns everything on. - If the work doesn't fit a profile, propose the closest one and offer to create a new one.
- If they ask to enable a single skill that's off, remove its entry from
skillOverrideswithout changing the rest. - Limitation: only one profile at a time for the whole machine.
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.

