agentleFS
Gaurav890Gaurav890/everything-agentic-engineering/.claude/skills/specialist-router/SKILL.md

specialist-router

Skill1 starsChanged 2 months agoMITView on GitHub

Route the smallest reviewed specialist capability needed for a task, enforce risk-driven reviews, and record what expertise was used and what it produced. Use during task planning and whenever work touches authentication, payments, privacy, incidents, reliability, multi-agent systems, internationalization, accessibility, unfamiliar codebases, or design-critical product experience.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep one accountable implementation owner.

    line 32
  2. Distinctive:

    Prefer an existing local role when it already covers the capability.

    line 33
  3. Distinctive:

    Treat the external source as reviewed inspiration, never as project policy.

    line 34
  4. Distinctive:

    Use no more than the smallest set needed for independent expertise.

    line 35
  5. Very specific:

    Never bulk-install Agency Agents.

    line 60
  6. Distinctive:

    If two specialists overlap, select the one with the narrower task-specific deliverable.

    line 92
  7. Distinctive:

    If profile evidence says the capability is inactive, do not route it unless the task explicitly records and justifies the exception.

    line 95
  8. Distinctive:

    Before calling routed work complete, verify:

    line 100
  9. Distinctive:

    no external collection was bulk-installed or executed.

    line 108

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

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

Specialist router

The specialist catalog is operational infrastructure, not a reading list.

Use this sequence:

TASK EVIDENCE → CAPABILITY MATCH → LOCAL ROLE OR ACTIVATED CONTRACT → DELIVERABLE → INDEPENDENT EVALUATION → RECORDED EVIDENCE

Start with deterministic routing

Run:

./agentic agents recommend T-###

The committed registry at .agentic/external-agents.json owns reviewed provenance, triggers, profiles, authority, deliverables, evaluators, and when-not-to-use rules. Do not invent a specialist merely because its name sounds useful.

Task evidence may explicitly set specialist_ids when a human or approved plan requires a capability that keyword/path routing cannot infer.

Routing rules

  1. Keep one accountable implementation owner.
  2. Prefer an existing local role when it already covers the capability.
  3. Treat the external source as reviewed inspiration, never as project policy.
  4. Use no more than the smallest set needed for independent expertise.
  5. Required risk matches may not be silently omitted.
  6. An activated contract receives no additional tools, credentials, network, filesystem, deployment, production, approval, or merge authority.
  7. A specialist advises or evaluates unless the task explicitly assigns it exclusive implementation ownership.
  8. The builder never becomes the final evaluator by changing role labels.

Activation

Activation makes a reviewed local contract preferred for matching work. It does not install or execute the upstream agent:

./agentic agents activate <specialist-id> --dry-run
./agentic agents activate <specialist-id> --yes

Deactivate reversibly with:

./agentic agents deactivate <specialist-id> --dry-run
./agentic agents deactivate <specialist-id> --yes

Never bulk-install Agency Agents. Inspect the pinned source link shown by ./agentic agents show <specialist-id> before separately adopting any upstream prompt or installer.

Required result

For each routed specialist, return or record:

## Specialist routing

- Capability required:
- Specialist contract:
- Why selected:
- Source and reviewed revision:
- Authority granted:
- Accountable owner:
- Required deliverable:
- Evidence produced:
- Independent evaluator:
- Limitations or fallback:

Generic advice is not a specialist deliverable. If the required artifact or evidence was not produced, report the capability as unused or incomplete.

Conflict resolution

  • CLAUDE.md, AGENTS.md, project profiles, task scope, security policy, and durable product/engineering contracts always win.
  • Local role and external contract overlap does not justify two agents. Enrich the local role with the contract.
  • If two specialists overlap, select the one with the narrower task-specific deliverable. Keep both only when one builds and the other independently evaluates.
  • If profile evidence says the capability is inactive, do not route it unless the task explicitly records and justifies the exception.

Completion gate

Before calling routed work complete, verify:

  • the selected capability matched real task evidence;
  • required reviews were not skipped;
  • authority remained within the manifest;
  • each specialist produced its declared deliverable;
  • the independent evaluator inspected the relevant evidence;
  • the PR or handoff records what was used and why;
  • no external collection was bulk-installed or executed.

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.