by Bil0000
A blank, working source of truth for Claude Code and Codex across macOS and Linux.
# Add to your Claude Code skills
git clone https://github.com/Bil0000/agents-fleetLast scanned: 9/1/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-09-01T09:12:34.322Z",
"npmAuditRan": true,
"pipAuditRan": true,
"promptInjectionRan": true
}agents-fleet is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by Bil0000. A blank, working source of truth for Claude Code and Codex across macOS and Linux. It has 50 GitHub stars.
Yes. agents-fleet passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/Bil0000/agents-fleet" and add it to your Claude Code skills directory (see the Installation section above).
agents-fleet is primarily written in Python. It is open-source under Bil0000 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh agents-fleet against similar tools.
No comments yet. Be the first to share your thoughts!
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
A blank, working source of truth for Claude Code and Codex across macOS and Linux, with optional multi-account failover.
Fleet keeps your own rules, settings, skills, commands, hooks, plugins, MCP servers, secrets, updates, and Claude-Mem setup consistent. This repository contains the engine and empty folders. It does not contain my own configuration.
Do not fork this repository. A fork of a public repository is public. Your Fleet will contain personal rules and device details.
origin must point to your repository, not Bil0000/agents-fleet.Set up this Agents Fleet in the repository I own. Before writing anything, verify that git remote get-url origin does not point to Bil0000/agents-fleet. Read SETUP_WITH_AN_AGENT.md in full and continue until every required check passes. Start with one device and local Claude-Mem unless I tell you that I have more devices. Preserve my current setup. Never put secrets, OAuth state, databases, or session history in Git.
The agent will back up your current setup, fill only the files you choose, install Claude-Mem, push to your private repository, wait for the checks, apply Fleet, and enable updates. The full contract is in SETUP_WITH_AN_AGENT.md.
If you do not want a GitHub repository, run git remote remove origin in a local clone and use manual apply. Automatic cross-device updates need your own remote repository.
rules/ shared, Claude-only, and Codex-only rules
settings/ safe client settings
skills/ shared skills
commands/ slash commands and prompts
hooks/ shared hook scripts
hosts/ one private overlay per device
mcp/ matching Claude and Codex MCP manifests
plugins.json shared plugin manifest
memory.example.toml optional shared-memory example
All user-owned files start empty. Add only what you want Fleet to manage.
main.live branch.live every minute and applies it safely.GitHub cannot directly enter devices behind a firewall. The one-minute poll gives the same result without opening an inbound port.
Your repository is independent from this template. Future template changes do not enter your Fleet automatically.
Run an update now:
~/.local/bin/fleet update
Inspect local additions:
~/.local/bin/fleet scan
Fleet can route any number of Claude and Codex subscriptions through one shared client home. Only the account that pays for a request changes. Threads, transcripts, memory, rules, settings, models, reasoning options, and native client behavior stay shared and provider-owned.
Set it up once on each macOS or Linux device:
fleet account setup
Fleet downloads and verifies CLIProxyAPI 7.2.149, installs a small retry bridge, and keeps both endpoints and their private keys on localhost. No separate proxy install or sudo access is needed. If Fleet finds a proxy setup that it does not own, it stops without changing it.
Use this only with accounts you own and where each provider's terms and plan allow it.
Open a new terminal, then use the normal provider commands to add or refresh accounts:
codex login
claude auth login
fleet account list
The same native logout commands remove the matching routed account. Logging in to the same account refreshes it instead of creating a duplicate. Fleet labels can be changed at any time.
fleet account rename codex ACCOUNT "Work Pro"
fleet account use codex "Work Pro"
fleet account auto codex on
fleet account current codex
Auto mode defaults to limits only: it switches accounts only for a structured subscription-limit response before output starts. fleet account auto codex on --when any also switches for temporary provider errors and can spend fallback quota. The retry bridge caps each wait at five minutes. The bundled provider-routing patch also caps quota cooldowns and clears them from fresh provider usage data. Fleet returns to the chosen primary when it is ready. Fleet never replays a request after output starts because that could repeat tool work or other side effects.
The same commands work with claude. Run fleet account auto --help to see the switching modes. fleet account current reports the account used by the latest successful 10-minute request bucket. Use fleet account enroll only to adopt credentials from an existing CLIProxyAPI setup.
Fleet does not contain a model list. New models, aliases, reasoning controls, auth screens, and non-auth client commands continue to come directly from Claude and Codex.
Newer Fleet-managed CLIProxyAPI builds are supported. The macOS updater source and provider-routing patch are included for existing custom builds. They are not installed by fleet account setup. The updater needs Go, a C compiler, and an existing Fleet provider-router setup. It checks a new build and restores the old build if activation fails.
After you create your own private repository, add a device profile:
./bin/fleet host add my-laptop --os macos --role workstation
Use --os linux for a Linux device. This adds a host manifest and the Claude-Mem plugin entry. It does not apply settings.
Commit and push the new profile to your private repository, then wait for its checks to pass and its live branch to update. For a full setup on that device:
./bin/fleet onboard --host my-laptop --memory local
This checks the tools and tests, applies Fleet, sets up account routing, and enables updates. It needs both Claude and Codex, Bun, Node.js 20.12 or newer, and a working private origin. Use the manual setup guide if you want only one client or no account routing.
The template leaves settings empty. In your private repo, you can add these to settings/codex.toml:
[features]
context_management = true
[agents]
max_concurrent_threads_per_session = 64
Use a Codex version that supports these settings. Fleet validates the thread limit as a positive whole number. It keeps your other agent roles and settings.
Shared hooks can go in settings/codex-hooks.json. Fleet combines them with hosts/HOST/codex-hooks.json. Shared and host Claude hooks are also combined.
Local Claude-Mem is still the default. For shared memory, copy memory.example.toml to memory.toml in your private repo. Set one Linux server and a client_hosts list of macOS or Linux clients. The older single client_host field also works. Use only one of these fields.
The shared setup uses Claude-Mem 13.24.0 and checks its worker before applying the known session and health fixes. Set up Fleet account routing on the server first. Clients connect through SSH. See the shared-memory setup steps.
fleet apply --dry-run shows targets before a write.systemd-creds on Linux../bin/fleet check
./bin/fleet apply --dry-run --host HOST
./bin/fleet apply --host HOST
~/.local/bin/fleet status --host HOST
~/.local/bin/fleet enable-updates
~/.local/bin/fleet update --host HOST
~/.local/bin/fleet scan --host HOST
~/.local/bin/fleet rollback
~/.local/bin/fleet auth set SECRET_NAME
~/.local/bin/fleet account setup
~/.local/bin/fleet account list
~/.local/bin/fleet account current
~/.local/bin/fleet account use codex ACCOUNT
~/.local/bin/fleet account auto codex on
systemd-creds and passwordless sudo -n for Fleet-managed secrets on Linux