fleetwright

A work-list system for coordinating several AI agents on separate items, such as files, pages, or records.

In plain words
What is it for?
Use it to split a list among three or more agents, track claimed work, and resume items left by a failed worker.
Why use it?
It prevents agents from taking the same item and gives each worker the same definition of what finished work means.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/narimannemo/fleetwright/skill
Any agent
npx skills add narimannemo/fleetwright --skill skill
Clone the repo
git clone --depth 1 https://github.com/narimannemo/fleetwright

Made for: Claude Code, Codex.

Per session 91 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,139 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00091 $0.02139
Opus 5 $0.00046 $0.01069
Sonnet 5 $0.00018 $0.00428
Haiku 4.5 $0.00009 $0.00214

Measured yesterday against content hash 61fc996d0238, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

fleetwright scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

src/fleetwright/skill/SKILL.md · 215 lines

How it starts

The opening of the file, as written. The whole thing — 215 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Running a fleet of subagents

When you spawn ten subagents on one job, two things go wrong and they have the same cause. A subagent you spawn has no context. It cannot see the other nine, and it did not read the reasoning you did before spawning it.

So they all start on the first item. And each one decides for itself what "done" means, so you get ten standards on one corpus.

Putting the task in the spawn prompt does not fix it. That prompt is invisible to the worker you spawn an hour later, and to the one that picks up work a crashed worker dropped.

fleetwright is the shared list they claim from, and the brief travels with each unit. You define the work once; every worker gets it at claim time.

When to use this

Use it when the user wants 3 or more subagents over a list of independent units.

Do not use it for one agent, for two units, or when each step depends on the last. A plain loop is clearer and the setup is not free.

First, in any session: find out where things are

You have no memory of previous sessions. Before deciding anything, run:

fleetwright state

It finds the database even if you do not know its name, and tells you which runs exist, which are still going, what failed, and the single next command. If it says there is no database here, this project has not used fleetwright and you are starting fresh.

If a run is still going, do not start a second one over the same work. Join it: spawn workers against the same database and they will claim what is left.

The whole flow

1. Set up the queue (you, once)

DB=$PWD/work.db

# What this run is. Everything below belongs to it.
RUN=$(fleetwright start --db "$DB" --label "extract claims from tomus II")

# What the work IS. The ninth worker, spawned an hour from now, reads this.
fleetwright define extract --db "$DB" \
  --instructions 'Read $path. Record every claim it makes, quoting verbatim.' \
  --done-when    'every claim in the file is recorded, or you have established there are none' \
  --returns      '{"claims": <int>, "notes": "<string>"}'

# The units. Use ABSOLUTE paths: workers may not share your directory.
ls scans/*.png | xargs -n1 basename > units.txt
fleetwright add extract --db "$DB" --from-file units.txt --run "$RUN" \
  --meta "{\"path\": \"$PWD/scans/\$name\"}"

Read the full file on GitHub · 215 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 215 lines · 91 tokens per session scan A 61fc996d0238

Subscribe to this mod's changes

fleetwright is a skill published in the GitHub repository narimannemo/fleetwright (0 stars, last pushed 24d ago), licensed Apache-2.0. It adds 91 tokens to every session and 2,139 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

rn-feature-development

This skill should be used when building any new feature in a React Native or Expo app, or when the /rn-dev-agent:rn-feature-dev command runs. Triggers on "build a feature", "add X to the app", "implement Y", "create a new screen", "rn-feature-dev", "feature development", "build me X", "add a screen", "wire up this…

Lykhoyda/rn-dev-agent · 86 tokens

rn-device-control

This skill should be used when the user asks to "control the simulator", "take a screenshot", "boot the emulator", "install the app", "read UI hierarchy", "manage device state", "open a deep link", "grant permissions", "stream native logs", "disable animations", "change device locale", or needs guidance on xcrun…

Lykhoyda/rn-dev-agent · 98 tokens

rn-testing

This skill should be used when the user asks to "write a Maestro test", "create E2E flows", "add testIDs", "run UI tests", "run E2E tests", "verify a feature works", "test my screen", "set up maestro-runner", "mock network requests", "inspect store state", "write test assertions", or needs guidance on test timing…

Lykhoyda/rn-dev-agent · 106 tokens

using-rn-dev-agent

Entry point for the rn-dev-agent plugin. Maps user intent to the right command, agent, or skill. Use at the START of any React Native development conversation. Triggers on "I want to build", "build a feature", "add a feature to the app", "test this", "something is broken", "fix the crash", "help with my React Native…

Lykhoyda/rn-dev-agent · 92 tokens

capturing-proof

This skill should be used when the user asks to "capture proof", "record a demo of this feature", "make a video showing it works", "record the flow for the PR", "generate a PR body", "capture screenshots for the PR", "proof-capture", or when a verified feature needs PR-ready proof artifacts (video + numbered…

Lykhoyda/rn-dev-agent · 105 tokens

creating-actions

This skill should be used when the user asks to "create an action", "save this flow as an action", "make this replayable", "record a reusable action", "author a Maestro flow as an action", "add a login/setup action", or when a verified UI walk should be persisted under .rn-agent/actions/ so future sessions can replay…

Lykhoyda/rn-dev-agent · 127 tokens