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.
npx agentmods add instructions/mrbenj/fleet-commander/claude-mdgit clone --depth 1 https://github.com/MrBenJ/fleet-commanderWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/mrbenj/fleet-commander/claude-md)<a href="https://agentmods.dev/instructions/mrbenj/fleet-commander/claude-md"><img src="https://agentmods.dev/badge/instructions/mrbenj/fleet-commander/claude-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.02884 | $0.02884 |
| Opus 5 | $0.01442 | $0.01442 |
| Sonnet 5 | $0.00577 | $0.00577 |
| Haiku 4.5 | $0.00288 | $0.00288 |
Grade A, and why
fleet-commander CLAUDE.md 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 4d ago.
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.
How it starts
The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
For agent-neutral guidance (self-contained), see AGENTS.md. Keep both files in sync when architecture changes.
Build & Run
# Build the binary
go build -o fleet ./cmd/fleet/
# Run directly without building
go run ./cmd/fleet/ <command>
# Install to PATH
go install ./cmd/fleet/
# Run tests
go test ./...
# Run a single package's tests
go test ./internal/monitor/...
# Build web UI + Go binary with embedded SPA (installs to PATH)
make build-all
# Build web UI only
make build-web
# Run web tests
cd web && npm run test
The binary must be placed alongside fleet.tmux.conf for the tmux config to be auto-sourced (it looks for the conf file relative to the executable path).
Tech stack: Go 1.21+ (backend), React 18 + TypeScript + Vite (frontend), tmux (session management), git worktrees (isolation).
Architecture
Fleet Commander is a CLI + TUI tool for managing parallel AI coding agent sessions, each in its own git worktree. The user stays in control — there is no AI coordinator.
Data model:
Fleet— ownsRepoPath,ShortName, and a list ofAgents. Persisted at.fleet/config.jsoninside the managed repo.Agent—Name(unique, used as tmux session:fleet-<Name>),Branch,WorktreePath,Driver(one ofclaude-code,codex,aider,kimi-code,antigravity,generic),StateFile, optionalPersona,FightModebool.- Global index:
~/.fleet/repos.jsontracks all registered repos for cross-repo commands (fleet queue --all).
Package layout:
cmd/fleet/— Cobra CLI wiring split acrossmain.go,cmd_agents.go,cmd_context.go,cmd_launch.go,cmd_repos.go,cmd_utils.go. Commands:init,add,list,start,attach,stop,queue,remove,hint,hangar,context,repos,launch,signal,unlock,rename,clear.cmd/fleet/embed.go—go:embed all:webdistdirective for the compiled React SPA (cmd/fleet/webdist/).internal/fleet/— Core data model.FleetandAgentstructs, JSON config persistence at.fleet/config.json.Load()walks up the directory tree to find a fleet.internal/global/— Global cross-repo state at~/.fleet/(repos index, global context/log).internal/tmux/— Thin wrapper around tmux CLI commands. Sessions are namedfleet-<agentName>. Sourcesfleet.tmux.confafter creating a session (best-effort).internal/monitor/— Reads tmux pane content viacapture-paneand classifies agent state asworking,waiting,stopped, orstarting. State detection inspects only the bottom 15 lines and checks for Claude Code-specific patterns: "esc to interrupt", spinner characters, "Esc to cancel",❯prompt, etc.internal/tui/— Bubble Tea TUI forfleet queue. Lists all agents with live state indicators, refreshes every 2 seconds. Selecting an agent starts its session if stopped, then attaches. After detaching (Ctrl+B, Q), the TUI loop restarts automatically.internal/worktree/— Git worktree CRUD:Create,CreateFromExisting,Remove,List,Move,Exists.Manager.Create(path, branch)runsgit worktree add -b <branch> <path>.internal/context/— Shared context store (.fleet/context.json). Agents can publish and read context from each other. Uses POSIX advisory file locks (flock) on.fleet/context.lockwith a 5-second timeout. Supports named private channels with fixed membership (names are used as-is; an empty name defaults tochannel-default).internal/state/— State file read/write for agent state signaling between Claude Code hooks and the monitor. Atomic writes via temp-file + rename. States live in.fleet/states/<name>.json.IsStale(ttl)helper for fallback-detection.internal/hooks/— Claude Code hook injection/removal. Injectsfleet signal working(PreToolUse) andfleet signal waiting(Stop) into.claude/settings.jsonin each worktree. Entries are tagged with"_fleet": truefor safe removal. Hooks readFLEET_STATE_FILEandFLEET_AGENT_NAMEenv vars.internal/hangar/— Web-based squadron mission control (fleet hangar). Go HTTP server + embedded React SPA.internal/hangar/server.go— HTTP server, route registration, SPA serving (embedded or Vite dev proxy).internal/hangar/tui.go— Bubble Tea TUI shown in the terminal while the server runs (zen/log views).internal/hangar/api/— REST handlers for fleet info, personas, drivers, squadron launch, agent generation, and stop.internal/hangar/ws/— WebSocket hub that polls.fleet/context.jsonevery 2s and broadcasts new channel messages and agent state changes.internal/hangar/terminal/— WebSocket-to-PTY proxy fortmux attach-session, powers the browser terminal.
internal/driver/— Driver interface and implementations (claude-code,codex,aider,kimi-code,antigravity,generic). Each driver implements state detection, hook injection (where supported), and command building.internal/squadron/— Squadron launch logic.RunHeadless()creates worktrees, tmux sessions, context channels, and prompt files.BuildMergerSuffix(),BuildConsensusSuffix(),ApplyPersona(), and fight-mode suffix assembly live here.web/— React 18 + TypeScript + Vite SPA for the hangar UI. Built withmake build-web, embedded into the Go binary viamake build-all.web/src/components/wizard/— Squadron setup wizard (SetupStep → AgentsStep → ReviewStep, with PersonaStep as a per-agent modal from AgentsStep).web/src/components/mission/— Mission control view (MissionControl, ContextLog, AgentPill, AgentTooltip, MultiView).web/src/components/terminal/— xterm.js terminal page for "Assume Control".web/src/hooks/—useWebSocket(auto-reconnect JSON event stream),useFleet(fleet data fetching).web/src/api.ts— Typed API client for all REST endpoints.
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.
- 4d ago First seen · 133 lines · 2,884 tokens per session scan A cdb1598a8443
fleet-commander CLAUDE.md is an instructions file published in the GitHub repository MrBenJ/fleet-commander (11 stars, last pushed 1mo ago), licensed MIT. It adds 2,884 tokens to every session, about $0.0144 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 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.