agentleFS
Sign inStart free
zakamezakame/brooks-agent-team/skills/assemble-with-fleet/SKILL.md

assemble-with-fleet

Skill4 starsChanged 50 days agoMITView on GitHub

Copilot CLI fleet mode or OpenCode task tool — Spawn a Brooks Surgical Team using parallel subagents — one independent agent per role, working in parallel with a shared task list. Skips orientation and goes straight to spawning.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 7 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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
  1. Distinctive:

    When invoked, do the following steps in order.

    line 16
  2. Distinctive:

    Read just enough to know what work is ahead:

    line 20
  3. Very specific:

    Do NOT do a full orientation — that is the assemble-team skill's job.

    line 25
  4. Distinctive:

    Ask the user one focused question:

    line 30
  5. Distinctive:

    Wait for the user's response before continuing.

    line 42
  6. Distinctive:

    Copilot CLI: use the session SQL database.

    line 48
  7. Very specific:

    Add dependency links so [review] and [test] tasks are blocked until their corresponding [implement] task is complete.

    line 57
  8. Distinctive:

    OpenCode: do NOT rely on the session SQL database.

    line 60
  9. Very specific:

    Copilot CLI: use mode: "background" and any fleet-specific fields required.

    line 68
  10. Very specific:

    OpenCode: do not use mode. Always include subagent_type.

    line 69
  11. Very specific:

    Do NOT ask the user to copy-paste anything or run /fleet manually — spawn the agents yourself.

    line 72
  12. Distinctive:

    After spawning, tell the user:

    line 146

≈ “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.

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

Assemble Team with Fleet Mode

<SUBAGENT-STOP> You are already operating as a dispatched subagent. Do not invoke `assemble-with-fleet`, `assemble-with-grok-team`, `assemble-with-hermes-team`, `assemble-with-pi-team`, `using-brooks-team`, or `surgeon` skills. Invoking an assembly skill from inside a subagent would spawn additional subagents recursively. </SUBAGENT-STOP><CODEX-STOP> OpenAI Codex has no fleet mode or task tool, so this skill does not apply there. If you are running under Codex, stop here and use explicit custom-agent dispatch instead: prompt the main session to spawn the specialist agents defined in `.codex/agents/` (see `AGENTS.md` for role routing). </CODEX-STOP>

When invoked, do the following steps in order.

Step 1: Quick Project Survey

Read just enough to know what work is ahead:

  • Project name and primary language/framework
  • Any open tasks, in-progress work, or recent git activity (git log --oneline -5)
  • Presence of AGENTS.md, COPILOT.md, docs/, tests/ directories

Do NOT do a full orientation — that is the assemble-team skill's job. This skill is about getting to work fast.

Step 2: Determine Which Roles to Spawn

Ask the user one focused question:

I'll spawn the core surgical team now. The default is Surgeon (you) + Copilot + Tester.

Do you want any additional teammates?

  • Editor — for documentation-heavy work
  • Toolsmith — if you need new scripts or automation
  • Language Lawyer — if tricky framework/version edge cases are expected
  • Program Clerk — if the codebase needs reorganization alongside new work

Reply with the roles to add, or say "default" to proceed with the core three.

Wait for the user's response before continuing.

Step 3: Build the Shared Task List

Using the project survey from Step 1, construct an initial task list.

Copilot CLI: use the session SQL database. Insert tasks with appropriate tags:

  • [implement] — Surgeon claims these
  • [review] — Copilot claims these (depend on corresponding [implement] tasks)
  • [test] — Tester claims these (depend on corresponding [implement] tasks)
  • [docs] — Editor claims these (if Editor role was requested)
  • [tooling] — Toolsmith claims these (if Toolsmith role was requested)
  • [research] — Language Lawyer claims these (if Language Lawyer role was requested)
  • [structure] — Program Clerk claims these (if Program Clerk role was requested)

Add dependency links so [review] and [test] tasks are blocked until their corresponding [implement] task is complete.

OpenCode: do NOT rely on the session SQL database. Maintain the shared task list in the main thread and include the current task list (with dependencies) in each subagent prompt so they can claim work without database access.

Step 4: Spawn the Team

Using the roles confirmed in Step 2, spawn each teammate directly via the task tool.

Copilot CLI: use mode: "background" and any fleet-specific fields required. OpenCode: do not use mode. Always include subagent_type. If resuming an existing background agent, include both task_id and subagent_type together.

Do NOT ask the user to copy-paste anything or run /fleet manually — spawn the agents yourself.

File ownership to include in each agent's prompt (assign based on your project survey):

  • Copilot: read-only review access across all files; never edits source directly
  • Tester: owns tests/ and any *.test.* / *.spec.* files
  • Editor: owns docs/, README.md, and any *.md documentation files
  • Toolsmith: owns scripts/, Makefile, and any tooling/config files
  • Language Lawyer: read-only research; communicates findings, does not edit files
  • Program Clerk: proposes structure changes for Surgeon approval before executing

Copilot (always spawn)

Copilot CLI: use agent_type: "copilot" (this is a configured custom agent). OpenCode: use subagent_type: "copilot".

Prompt:

You are the Copilot on the surgical team for [PROJECT NAME]. Claim tasks tagged [review] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Ask the Surgeon for spec or plan context before reviewing anything. Do not edit source code — report findings only. File ownership: read-only access across all files.

Tester (always spawn)

Copilot CLI: use agent_type: "tester" (this is a configured custom agent). OpenCode: use subagent_type: "tester".

Prompt:

You are the Tester on the surgical team for [PROJECT NAME]. Claim tasks tagged [test] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Enumerate failure modes before writing any test code. Do not modify production code. File ownership: owns tests/ and any *.test.* / *.spec.* files.

Optional roles (spawn only if user requested them)

Language Lawyer (Copilot CLI: use agent_type: "language-lawyer"; OpenCode: use subagent_type: "language-lawyer")

Editor, Toolsmith, Program Clerk (Copilot CLI: use agent_type: "general-purpose"; OpenCode: use subagent_type: "general")

Editor prompt:

You are the Editor on the surgical team for [PROJECT NAME]. Invoke the editor skill at the start of this session and before every task you claim. Claim tasks tagged [docs] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Read the code before writing about it. Never document what you haven't verified. File ownership: owns docs/, README.md, and any *.md documentation files.

Toolsmith prompt:

You are the Toolsmith on the surgical team for [PROJECT NAME]. Invoke the toolsmith skill at the start of this session and before every task you claim. Claim tasks tagged [tooling] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Every tool you build must have --help text and tests. File ownership: owns scripts/, Makefile, and any tooling/config files.

Language Lawyer prompt:

You are the Language Lawyer on the surgical team for [PROJECT NAME]. Invoke the language-lawyer skill at the start of this session and before every task you claim. Claim tasks tagged [research] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Cite the spec or docs — never guess. Say "I don't know" when uncertain. File ownership: read-only research; communicate findings, do not edit files.

Program Clerk prompt:

You are the Program Clerk on the surgical team for [PROJECT NAME]. Invoke the program-clerk skill at the start of this session and before every task you claim. Claim tasks tagged [structure] from the shared task list (Copilot CLI: session SQL database; OpenCode: main-thread task list in this chat). Propose all reorganizations to the Surgeon for approval before executing. File ownership: proposes structure changes; executes only after Surgeon approval.

Spawn all agents in a single response using parallel task tool calls.

Step 5: Confirm and Hand Off

After spawning, tell the user:

Team is live. You are the Surgeon — claim [implement] tasks and your teammates will handle review, testing, and support in parallel.

A few reminders:

  • Two teammates must not edit the same file — follow the ownership assignments above
  • Run the assemble-team skill at any time for a full role reference
  • Copilot CLI: use /tasks to monitor background teammates; OpenCode: use ctrl-x + down
  • When the work is done, cancel any idle background tasks and clean up team resources

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