agentleFS
bakhod1rbakhod1r/awesome-agents/.claude/skills/agent-registry/SKILL.md

agent-registry

Skill0 starsChanged 3 months agoMITView on GitHub

Look up which agent or team owns a kind of work in this repository. Use when deciding whom to delegate to, when the user names a role, or when asked "who handles X".

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the candidate's Decision Rules section.

    line 22
  2. Distinctive:

    If nothing fits, say so plainly and name the roster gap.

    line 58

“Common” is measured: kept by at least 6 in 10 of the 4 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

Agent Registry

Where the roster lives

FileContents
.claude/teams/README.mdEvery agent: title, name, team, model
.claude/teams/<team>.mdOne team: mission, charter, roster with missions
.claude/agents/<name>.mdOne agent: all ten sections

The name in frontmatter is what the Agent tool takes. It is the kebab-case slug.

Lookup procedure

  1. Reduce the request to a domain noun: "migration", "drift", "egress cost", "screen reader".
  2. Grep the roster for it. .claude/teams/README.md holds all 135 missions in one file.
  3. Read the candidate's Decision Rules section. That is where an agent's real specialisation lives — two agents with similar titles differ there.
  4. Pick one. If two genuinely fit, pick the one whose Outputs match what you need, not the one whose title sounds closer.

Disambiguation — the pairs people confuse

QuestionAnswerNot
Design the schemadatabase-architectdatabase-engineer (tunes and migrates an existing one)
Build a metrics pipelineobservability-engineerobservability-test-engineer (verifies signals exist and are correct)
Write the modelai-engineermlops-engineer (ships, versions, monitors it)
Is the model still good?model-monitoring-engineerai-evaluation-engineer (builds the eval sets and metrics)
Cloud account hardeningcloud-security-engineerplatform-engineer (builds on it)
Corporate servers, patching, IAMsystems-administrator, iam-engineerplatform-engineer (product infrastructure)
Attack itpenetration-tester (authorised, proves exploitability)security-test-engineer (in-CI security regression)
Watch the alertssoc-analystincident-response-engineer (commands the response)
Cost of a designcloud-cost-architectfinops-engineer (attributes and optimises a running bill)
Requirementsbusiness-analyst (process and rules)product-manager (what to build and why)

Worked example

"Our S3 bill tripled last month."

  • Domain noun: cost, running system, already spent.
  • finops-engineer — attribution and waste on an existing bill. First choice.
  • cloud-cost-architect — only if the finding is architectural (egress pattern, storage class by design). Escalate to it after attribution, not before.
  • data-engineer — pulled in if the driver turns out to be pipeline file-sizing.

Answer: finops-engineer first. Name the follow-ons as conditional, not parallel.

Rules

  • One agent per concern. Do not fan out to five for a one-file change.
  • Architects decide, engineers implement, quality verifies. Do not skip the middle.
  • If nothing fits, say so plainly and name the roster gap. Do not force the nearest agent — a wrong specialist is more confidently wrong than a generalist.

Done when

  • A single best-fit agent is named, with the name value, not just the title.
  • The reason for the fit cites the agent's Decision Rules or Outputs, not its title.
  • Near-misses are named as conditional follow-ons, not as parallel work.
  • A genuine gap is reported as a gap.

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.