← Back to Skills Marketplace
81
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install nm-conjure-agent-teams
Description
Coordinate Claude Code Agent Teams through filesystem-based protocol. Use
Usage Guidance
This skill coordinates local 'claude' CLI processes via files and tmux panes. Before installing, be aware it will: (1) create and write files under ~/.claude/, (2) spawn and kill tmux panes and run the `claude` CLI, and (3) expects environment variables (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS, CLAUDECODE) and may require ANTHROPIC_API_KEY in some modes — but the registry metadata did not declare these requirements. If you intend to use it: (a) verify you have a local `claude` CLI and tmux, (b) run it in an isolated or disposable environment (container or throwaway user account) if you are uncomfortable with it modifying your home dir or processes, (c) back up any important data in your working directory, and (d) prefer human oversight for spawn/kill operations. If you need higher assurance, ask the publisher to update registry metadata to list required binaries and env vars and to justify why each is needed.
Capability Analysis
Type: OpenClaw Skill
Name: nm-conjure-agent-teams
Version: 1.0.0
The skill bundle provides a complex framework for orchestrating multiple Claude CLI processes ('teammates') using tmux and a filesystem-based coordination protocol. It includes high-risk capabilities such as spawning new CLI sessions, bypassing nested session guards, and automated process management (killing and restarting tmux panes) as described in 'spawning-patterns.md' and 'health-monitoring.md'. While these behaviors are aligned with the stated purpose of multi-agent coordination, the ability to manage a fleet of processes and manipulate system-level locks (~/.claude/teams/) represents a significant attack surface and high-privilege access that could be abused if the agent is subverted via prompt injection.
Capability Tags
Capability Assessment
Purpose & Capability
The name/description (coordinate Claude Code Agent Teams via a filesystem protocol) aligns with the instructions and modules: writing JSON to ~/.claude, fcntl-based inbox locking, tmux-based agent spawning, role/risk checks via leyline modules. The declared required config paths (night-market.delegation-core, leyline:damage-control, leyline:risk-classification) also align with references in the modules.
Instruction Scope
SKILL.md contains concrete runtime steps that create and modify files under ~/.claude/, spawn and kill tmux panes, and launch separate `claude` CLI processes. It also references environment variables (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS, CLAUDECODE, and optionally ANTHROPIC_API_KEY/ apiKeyHelper) and expects tmux and the `claude` CLI to be present — but the package metadata did not declare required binaries or env vars. The instructions are local-only (no external endpoints), but they give the skill broad ability to manage processes and filesystem state on the host.
Install Mechanism
Instruction-only skill with no install spec or code files, so nothing will be downloaded or written by an installer. This reduces supply-chain risk compared to an installer that fetches arbitrary code.
Credentials
Registry metadata lists no required environment variables, yet SKILL.md requires/mentions several env settings (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS, CLAUDECODE) and references possible need for ANTHROPIC_API_KEY for `--bare` mode. The skill also expects access to the user's home directory (~/.claude) and ability to spawn/kill tmux panes and run claude — these are powerful privileges that are proportionate to the stated purpose but should have been surfaced in metadata.
Persistence & Privilege
always:false (no forced permanent inclusion). The skill instructs spawning background processes (tmux panes) and atomically writing config and task files to ~/.claude. Autonomous invocation is allowed (default) — combined with the ability to spawn/kill panes and modify files, this increases blast radius if misused, but autonomous invocation alone is not an immediate red flag given the skill's purpose.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install nm-conjure-agent-teams - After installation, invoke the skill by name or use
/nm-conjure-agent-teams - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
**Initial release of agent-teams skill for Claude Code Agent multi-agent coordination.**
- Enables coordination of multiple Claude CLI agent processes using a filesystem-based protocol.
- Provides team management, messaging, task assignment with dependencies, and health monitoring—all without needing a database or network service.
- Supports roles (implementer, tester, reviewer, architect, researcher) with capability matrices and sizing rules for mission-level teams.
- Integration guidance provided for Claude Code and Conjure delegation plugins.
- Includes documentation for protocol architecture, tmux-based spawning, message workflow, and alternative worktree isolation for file safety.
Metadata
Frequently Asked Questions
What is Nm Conjure Agent Teams?
Coordinate Claude Code Agent Teams through filesystem-based protocol. Use. It is an AI Agent Skill for Claude Code / OpenClaw, with 81 downloads so far.
How do I install Nm Conjure Agent Teams?
Run "/install nm-conjure-agent-teams" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Nm Conjure Agent Teams free?
Yes, Nm Conjure Agent Teams is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Nm Conjure Agent Teams support?
Nm Conjure Agent Teams is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Nm Conjure Agent Teams?
It is built and maintained by athola (@athola); the current version is v1.0.0.
More Skills