agentsclimarketplace

Agent fleet control

Skill runsagents/agent-fleet-control

Coordinates parallel coding agents with worktree isolation: one writer per worktree, bounded task contracts, and merge receipts.

Install
npx -y skills add runsagents/agent-fleet-control

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 26 days oldThe repository was created 26 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Coordinate multiple coding agents with isolated Git worktrees, explicit path territory, verification evidence, and human-reviewed merge receipts. Trigger on "fleet new", "spin up a second agent", or multi-agent work.

SKILL.md

2.1 KB, 424 tokens by cl100k_base, as published. Nobody here has run it

Agent Fleet Control

Use this skill when the user says fleet new, asks to "spin up a second agent," or requests multi-agent work with writing agents.

Invariants

  1. One writing agent per worktree. Read-only reviewers may inspect, but must not edit, a writer's checkout.
  2. Every writer gets explicit repository-relative path territory before work begins.
  3. Verification must pass for the exact current commit before a receipt is created.
  4. Never merge unattended. This skill and bin/agent-fleet never push or merge.

Workflow

  1. Confirm the repository has an initial commit and the intended task is bounded.

  2. Convert the assignment into a task contract: agent name, task, allowed paths, observable acceptance criteria, and one verification command.

  3. Run:

    bin/agent-fleet new <agent> "<task>" \
  --path <repository-relative-path> \
  --acceptance "<criterion>" \
  --verify "<command>"

  4. Give the writer only the created worktree and its recorded contract. Do not let another writer edit that worktree.

  5. If territory collides, change the scope. Use --override only after explicit human approval and call out the recorded exception.

  6. When the writer stops, inspect the diff and run bin/agent-fleet verify <agent>.

  7. After independent review, record the reviewer and every known risk:

    bin/agent-fleet receipt <agent> \
  --reviewer "<identity>" \
  --risk "<risk or None reported>"

  8. Hand the receipt, diff, and commit to a human merge owner. The human chooses whether and how to merge.

Boundaries

  • Treat allowed paths as coordination metadata, not an operating-system sandbox.
  • Treat reviewer identity as an assertion, not authentication.
  • Treat verification commands as trusted repository code; they execute with the caller's local permissions.
  • Do not place credentials in task text, criteria, commands, or logs.

What ships with it: 15 files

59.7 KB alongside SKILL.md, 4 of them executable

bin/

fixtures/

policies/

schemas/

templates/

tests/

Keep looking

Skills are one crate of 326,984. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.