agentleFS
Sign inStart free
youssef-110youssef-110/skills-conflict/skills/skill-router/SKILL.md

skill-router

Skill1 starsChanged 32 days agoMITView on GitHub

Manages the automatic skill router that picks the best installed skills for every request and blocks overlapping ones. Use when the user asks which skill or plugin to use for something, why a skill was or was not picked, which installed skills overlap, duplicate or conflict, says they have too many skills, wants to always prefer one skill over another, wants to stop a skill from being suggested, or wants to turn the router on or off. Also use when a <skill-router> note looks wrong.

60 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use context-engineering, accessibility 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.
  • mrgoonieSkill

    context-engineering

    Master context engineering for AI agent systems. Use when designing agent architectures, debugging context failures, optimizing token usage, implementing memory systems, building multi-agent coordination, evaluating agent performance, or developing LLM-powered pipelines. Covers context fundamentals, degradation patterns, optimization techniques (compaction, masking, caching), compression strategies, memory architectures, multi-agent patterns, LLM-as-Judge evaluation, tool design, and project development.

    Also in 3 projects that use this

  • ElasticSkill

    accessibility

    Accessibility guidance for Kibana. Use this skill when working with or reviewing EUI components, resolving a11y-related (@elastic/eui) ESLint issues, and ensuring proper use of ARIA attributes, focus management, keyboard interactions, and accessible naming conventions.

    Also in 3 projects that use this

  • MicrosoftSkill

    caveman

    Ultra-compressed response style that reduces output token count while preserving technical accuracy, with intensity levels and auto-clarity safety rules

    Also in 4 projects that use this

  • SupabaseSkill

    copywriting

    Write or audit UI copy (buttons, labels, empty states, error messages, tooltips, form text) anywhere in the monorepo. Load it before shipping or reviewing any user-facing text — including when copy is incidental to the task, like a new feature that adds buttons, toasts, dialogs, or validation messages.

    Also in 3 projects that use this

  • GoogleSkill

    teach

    Teach the user a new skill or concept, within this workspace.

    Also in 3 projects that use this

  • GoogleSkill

    domain-modeling

    Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.

    Also in 3 projects that use this

  • +4 more from google

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 51 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All commands go through one script.

    line 14
  2. Very specific:

    After changing preferences, run route again on the user's example to show the new behaviour.

    line 48
  3. Very specific:

    When the note was wrong for a request, check route --json to see what matched.

    line 49
  4. Distinctive:

    Learn: Every Skill and Agent call is counted.

    line 66

“Common” is measured: kept by at least 6 in 10 of the 51 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill router

A hook already runs before every prompt. It scans every installed skill, plugin skill, slash command and subagent, scores them against the request, and adds a <skill-router> note naming the best picks and the overlapping ones to skip. Nothing has to be invoked for that to work. This skill is for inspecting and steering it.

All commands go through one script. The exact command is in this session's "skill-router is active" startup note (node "<path>/scripts/cli.js" <command>). If that note is missing, the script is ../../scripts/cli.js relative to this skill's base directory. Do not search the disk for it.

node "<path>/scripts/cli.js" <command>

What the user wants, and the command for it

RequestCommand
"which skill should I use for X?" / "why did it pick Y?"route "<their task, verbatim>" (add --json for scores)
"which of my skills overlap / conflict / are duplicates?"audit (or audit --out skill-audit.md for a file)
"always use X instead of Y"prefer X (accepts several ids, first wins)
"stop suggesting X" / "never use X"ignore X (also takes a whole plugin name)
"only use X as a last resort"avoid X
undo any of thoseunprefer X, unignore X, unavoid X
"turn the router off / on"off / on
"is it working?"doctor
"how many skills does it know about?" / "what do I use most?"stats
just installed or removed skillsnothing: the index refreshes itself; rebuild forces it
tune sensitivityconfig minConfidence 0.35 (higher = fewer notes), config maxPicks 2

Ids are exactly what the router prints, for example superpowers:brainstorming, taste-skill, engineering:debug. If an id is wrong, the command lists close matches. Use one of those instead of guessing.

Answering well

  • For "which skill for X", run route and explain the pick in a sentence. Name the overlapping skills it rejected and why. Do not paste raw JSON.
  • For an audit, lead with the 3 to 5 biggest overlap groups, say which member the router currently favours, and offer prefer commands for the user to accept. Do not set preferences the user did not ask for.
  • After changing preferences, run route again on the user's example to show the new behaviour.
  • When the note was wrong for a request, check route --json to see what matched. Usually one of these fixes it: prefer the right skill, ignore a noisy one, or raise minConfidence.

How it decides (for explaining to users)

  1. Index. Reads SKILL.md, command and agent frontmatter from project .claude/, ~/.claude/, every enabled plugin and Claude-desktop plugins. Only folders Claude Code really loads count, so uninstalled plugins are never suggested.
  2. Score. BM25 over names and descriptions. Task words such as "debug", "review" and "deploy" outweigh context nouns. Words in a skill's own name, the author's quoted trigger phrases, and explicit mentions ("use taste-skill") get large boosts.
  3. De-duplicate. Two candidates from different plugins whose descriptions are too similar (tf-idf cosine) do the same job. Only the higher-scoring one is kept and the other goes on the "do not also load" list. Skills from the same plugin are treated as a team, never as rivals.
  4. Decide. A confident match becomes a directive ("USE FIRST"). A weaker one becomes a short list for Claude to choose from. Chit-chat, "continue", slash commands and simple look-ups get nothing.
  5. Learn. Every Skill and Agent call is counted. Inside an overlap group, the skills you actually use win ties.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free