skill-routing
Skill4 starsChanged 3 months agoMITView on GitHub
Match a user request to the right skill using trigger matching, project context, and conversation history. Score ambiguity 1–10 and resolve it using context signals or a single clarifying question. Load when project-orchestrator needs to route a request, or when multiple skills could match and the right one is unclear. Also triggers on "which skill should handle this", "route this request", "I'm not sure which skill to use", "disambiguate this", "skill routing". Called by project-orchestrator for request routing. For finding whether a skill exists for a capability, use skill-finder instead.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- naman10parikh/pennytheir own version, first seen Nov 8, 2026
- naman10parikh/wikimemtheir own version, first seen Nov 8, 2026
- naman10parikh/turbotheir own version, first seen Oct 25, 2026
- naman10parikh/pixeltheir own version, first seen Oct 25, 2026
- naman10parikh/agentgridtheir own version, first seen Oct 20, 2026
- i-9-ai/skillstheir own version, first seen Sep 24, 2026
Used with
Projects with it also use ecosystem-reference, fractal-delegation 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.
naman10parikhSkillecosystem-reference
Quick-access directory of Claude Code ecosystem repos, skills, plugins, MCPs, and patterns. Use when improving the harness.
Also in 5 projects that use this
naman10parikhSkillfractal-delegation
Also in 5 projects that use this
sublimotionSkillagent-runtime
Launch and manage detached agent runs on EKS via the agent-runner CLI. Use when the user says "launch an agent run", "agent runtime", "agent-runner", "run this spec on the cluster", "check run status", "agent run logs", or "stop a run". Do NOT use for writing the agent-runner code itself (edit the sibling ../agent-runner repo), Terraform of the foundation (use terraform-automation), or GPU serving deployments (use infra-deployer).
Also in 5 projects that use this
GoZippySkillrepo-scout
Also in 5 projects that use this
BenjaminGSkillharness-review
Turn one PR's review comments into harness fixes — classify each cause by the most solid place it can live (codebase, lint, rules, skill), then draft the fixes in a separate PR.
Also in 5 projects that use this
botirkhaltaevSkilldesign-systems
Build and maintain scalable design systems: design tokens, component libraries, naming conventions, governance, and design-to-code pipelines. Covers Figma variables, token architecture, atomic design, component APIs, versioning, and team workflows. Use when creating a design system, scaling an existing one, or establishing design-engineering alignment. Triggers on: "design system", "design tokens", "component library", "token naming", "Figma variables", "atomic design", "design ops", "design-to-code".
Also in 5 projects that use this
- +6 more from naman10parikh
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- naman10parikh/agentgrid★ 9same name
- naman10parikh/wikimem★ 8same name
- i-9-ai/skills★ 0same name
- Kirchlive/skill-routing★ 0same name
- naman10parikh/penny★ 0same name
- naman10parikh/pixel★ 0same name
- naman10parikh/turbo★ 0same 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 repository125 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/agent-loom-routing.mdcCursor rule
- .cursor/rules/agent-loom-skills-index.mdcCursor rule
- adversarial-hatSkill
- agent-builderSkill
- agent-launcherSkill
- agent-loom-syncSkill
- agent-observabilitySkill
- agent-run-retroSkill
- agent-system-architectureSkill
- api-and-interface-designSkill
- api-deprecation-and-migrationSkill
- apply-paper-to-projectSkill
- app-security-hardeningSkill
- architectural-decision-logSkill
- assumption-mappingSkill
- brainstormingSkill
- browser-testing-with-devtoolsSkill
- business-modelingSkill
- ci-cd-and-automationSkill
- codebase-understandingSkill
- code-review-crspSkill
- code-simplificationSkill
- compress-skillSkill
- context-engineeringSkill
- create-agent-promptSkill
- cross-link-skillsSkill
- customer-discoverySkill
- debug-and-fixSkill
- deep-thinkingSkill
- dependency-mappingSkill
- deploy-anywhereSkill
- deprecate-skillSkill
- design-directionSkill
- design-reviewSkill
- design-systemSkill
- dynamic-routingSkill
- eval-judgeSkill
- eval-outputSkill
- eval-pipelineSkill
- eval-rubric-designSkill
- experimentationSkill
- experiment-backlogSkill
- experiment-readoutSkill
- experiment-runbookSkill
- experiment-specSkill
- fault-localizeSkill
- feature-specSkill
- fermiSkill
- first-principlesSkill
- frontend-designSkill
- generate-changelogSkill
- git-workflow-and-versioningSkill
- gsap-animationSkill
- harness-engineeringSkill
- harness-evolutionSkill
- harness-generationSkill
- idea-evaluationSkill
- idea-generationSkill
- implementation-planSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 6 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
6
Distinctive
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
- Distinctive: line 30
Never guess when ambiguity is ≥ 7: Ask exactly one disambiguation question.
- Distinctive: line 32
Always read skill descriptions before routing: Do not rely on skill names alone.
- Very specific: line 41
Do NOT use a hardcoded routing table.
- Very specific: line 49
Before scoring ambiguity, read references/harness-symptoms.md.
- Very specific: line 51
If the user describes agent misbehavior (not app bugs), prepend harness-engineering to candidates and lower ambiguity when manifest is …
- Very specific: line 54
If project-setup just completed in this session → route harness-generation next unless user opted out.
- Distinctive: line 73
Score 4–6: Read these context signals to resolve silently:
- Distinctive: line 81
Score 7–10: Ask ONE disambiguation question as a binary choice:
- Distinctive: line 88
Before returning the routed skill, verify its pre-req column:
- Very specific: line 112
When project-orchestrator calls you, return fast.
- Distinctive: line 169
No changes — citation audit passed.
“Common” is measured: kept by at least 6 in 10 of the 7 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 Routing
You are a Skill Router. Given a user request, project state, and conversation context, you determine which skill to invoke. When multiple skills match, you score the ambiguity and resolve it — either silently via context or by asking one targeted question.
Hard Rules
Never guess when ambiguity is ≥ 7. Ask exactly one disambiguation question. Never ask more than one question. Frame it as a choice between the top 2 candidates. Always read skill descriptions before routing. Do not rely on skill names alone. Always log the routing decision — skill chosen, ambiguity score, and how it was resolved.
Workflow
Step 1 — Discover Candidate Skills
Do NOT use a hardcoded routing table. Call skill-finder with the user's request to discover all candidate skills from the live library.
- 1 candidate → return it. Done. Ambiguity = 1.
- 0 candidates →
skill-finderfound nothing. Returnno-matchto the caller. - 2+ candidates → proceed to Step 2.
Step 1b — Harness symptom pre-route
Before scoring ambiguity, read references/harness-symptoms.md.
If the user describes agent misbehavior (not app bugs), prepend harness-engineering
to candidates and lower ambiguity when manifest is missing or symptoms match.
If project-setup just completed in this session → route harness-generation next unless user opted out.
Step 2 — Score Ambiguity (1–10)
Rate how unclear the routing is:
| Score | Meaning | Criteria |
| 1–3 | Low | One candidate clearly dominates — better trigger match, correct phase, or natural next step |
| 4–6 | Medium | Two candidates are plausible but context signals can resolve |
| 7–10 | High | Two+ candidates are equally plausible even after reading context |
Factors that RAISE the score: request uses generic words ("plan", "review", "improve"), no project artefacts to anchor phase, request is short/vague. Factors that LOWER the score: request uses a skill's exact trigger phrase, project phase clearly narrows candidates, conversation history shows a chain in progress.
Step 3 — Resolve Ambiguity
Score 1–3: Route to the best match. No question needed.
Score 4–6: Read these context signals to resolve silently:
- Project phase — what artefacts exist narrows the candidates
- Conversation history — what the user just finished suggests the next step
- Trigger precision — does the phrasing match one skill's triggers more exactly?
- Upstream/downstream — is one skill the natural successor of the last completed skill?
Pick the best fit. Note the runner-up in the routing log.
Score 7–10: Ask ONE disambiguation question as a binary choice:
"This could be [skill-A] ([one-line purpose]) or [skill-B] ([one-line purpose]). Which fits?"
Route based on the answer.
Step 4 — Check Pre-requisites
Before returning the routed skill, verify its pre-req column:
- Pre-req missing → tell the caller which skill produces the pre-req
- Pre-req exists → return the skill
Step 5 — Return Routing Decision
Return to the calling skill (usually project-orchestrator):
Routed: [skill-name]
Ambiguity: N/10
Candidates: [skill-a, skill-b, ...]
Resolved by: [exact-match | phase-context | conversation-chain | upstream-downstream | user-answer]
Pre-req: [met | missing — need X first]Gotchas
- "Review" is the most ambiguous word — "review this code" →
code-review-crsp, "review changes for context" → not a review skill, "review the plan" →adversarial-hat. Always check the object being reviewed. - "Plan" is similarly overloaded — "plan this change" →
problem-to-plan, "plan implementation" →implementation-plan, "plan this out" →process-decomposer. The verb before "plan" disambiguates. - Return only concrete, invokable skill names. Caller-owned labels like "Phase recommendation" are not valid routing results.
- A skill scoring high on trigger match but failing the pre-req check is NOT the right skill to invoke directly — route to the pre-req skill first.
- When
project-orchestratorcalls you, return fast. Routing should add seconds, not minutes. - Agent complaints are usually harness, not debug. "Agent keeps failing" →
harness-engineering, notdebug-and-fix, unless the user names a specific code bug.
Example
<examples>
<example>
<input>User says: "Help me think through this problem"</input>
<output>
Candidates: deep-thinking, brainstorming, socratic
Ambiguity: 5/10Context signals:
- Project phase: specs exist, PRD exists → past design phase
- Conversation: user just finished implementation, hit a wall
- Trigger match: "think through" matches deep-thinking and brainstorming equally
- Upstream: post-implementation thinking → deep-thinking (not brainstorming, which is pre-implementation)
Routed: deep-thinking
Ambiguity: 5/10
Candidates: deep-thinking, brainstorming, socratic
Resolved by: upstream-downstream (post-implementation context)
Pre-req: met
</output>
</example>
</examples>
Common Rationalizations
| Excuse | Reality |
| "Obvious which skill" | Obvious to user ≠ obvious to router at score 6/10. |
| "Invoke both skills" | Double invocation wastes tokens and causes conflicts. |
| "Skip ambiguity score" | Score drives whether to ask — skipping hides misfires. |
| "Library-first is dogma" | Library-first is default — project skills win when present. |
| "Route to general-purpose" | Named skills encode workflows general chat skips. |
Verification
- Ambiguity score 1-10 stated when multiple skills match
- Winner skill named with one-line rationale
- User asked to disambiguate when score ≥7
- Project-local skill preferred over global when both match
Red Flags
- Ambiguous review routed without disambiguating code vs UI
- Plan request sent to wrong planner skill for intent
- Non-invocable label returned instead of concrete skill name
- Multiple candidates returned without disambiguation step
Prune Log
Last pruned: 2026-07-04
- No changes — citation audit passed; content current (improve-skills full pass 2026-07-04)
Impact Report
Routing complete: [request summary]
Skill routed: [skill-name]
Ambiguity: N/10
Candidates considered: [list]
Resolved by: [method]
Question asked: [yes — question text | no]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.

