一键导入
project-manager-agent
Monitors active sub-agents, detects stalls, provides status updates, and steers stuck agents via sessions_send.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Monitors active sub-agents, detects stalls, provides status updates, and steers stuck agents via sessions_send.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Seamless bidirectional bridge between nanobot (Ollama Mistral orchestrator) and overstory (Claude Code agent swarm). Routes tasks through the OverClaw gateway (port 18800) to overstory for subagent coordination, syncs memory.
Integrates OpenClaw with the FunClip AI video editor to enable AI-powered video recognition and clipping via command line.
IMPORTANT: OpenRouter is required. Routes tasks to the right model and always delegates work through sessions_spawn. Rejects prompt-injection patterns in task strings (v1.7.6+).
Automatically create clips and videos from media files in a specified folder. Uses Agent Swarm for intelligent task delegation and supports cron-based scheduling.
Analyzes journal entries and chat history to identify high-value topics and automatically generate blog posts.
Extracts and analyzes Cursor IDE chat history to identify key discoveries, obstacles, and solutions, saving findings to the journal.
| name | project-manager-agent |
| description | Monitors active sub-agents, detects stalls, provides status updates, and steers stuck agents via sessions_send. |
Monitors active sub-agents, detects stalls, provides status updates, and steers stuck agents via sessions_send.
This skill provides oversight for your OpenClaw sub-agents. It proactively checks the status of running sub-agents, identifies potential stalls or failures, and can steer stalled sub-agents by sending a continue message via the sessions_send tool.
python3 <skill-dir>/scripts/project_manager.py
--staleness_threshold_minutes <int>: Minutes after which a sub-agent is considered stalled (default from config or 10).--check_all_sessions: Check all sessions, not just subagents of the current main session.--json: Print human summary then a JSON line with stalledSessionKeys, steerMessage, and summary for agent steering.--no_cooldown: Output all stalled keys (ignore 15-minute cooldown).--record-steered KEY [KEY ...]: Record these session keys as just steered (for cooldown). Run after calling sessions_send.python3 /Users/ghost/.openclaw/workspace/skills/project-manager-agent/scripts/project_manager.py --staleness_threshold_minutes 15
python3 /Users/ghost/.openclaw/workspace/skills/project-manager-agent/scripts/project_manager.py --json --staleness_threshold_minutes 10
Then parse the last line (JSON), call sessions_send for each key in stalledSessionKeys, and optionally run --record-steered <keys>.
stalledSessionKeys for steering.--json, parse the output and call sessions_send for each stalled session (see Agent instructions below).When the user or a cron asks you to run the project-manager-agent skill:
--json (and existing args as needed):
python3 /Users/ghost/.openclaw/workspace/skills/project-manager-agent/scripts/project_manager.py --json --staleness_threshold_minutes 10
stalledSessionKeys, steerMessage, and summary. Parse it (e.g. take the last line that starts with { and ends with }).stalledSessionKeys is non-empty:
stalledSessionKeys (cap at 5 per run to avoid spam), call sessions_send(sessionKey, steerMessage from JSON or "Please continue working on your task.", 0) (fire-and-forget; timeoutSeconds=0).python3 /Users/ghost/.openclaw/workspace/skills/project-manager-agent/scripts/project_manager.py --record-steered <key1> <key2> ...
You have access to sessions_list and sessions_send; the script only produces the list of session keys to steer. Steering is done by you in the same turn via the sessions_send tool.
See config.json for: staleness_threshold_minutes, steer_cooldown_minutes, steer_message, max_steer_per_run.
This skill is ideal for a cron job, running periodically (e.g., every 5 minutes). The agent runs the script with --json, steers stalled subagents via sessions_send, and announces the status.
Example Cron Job Payload:
{
"payload": {
"kind": "agentTurn",
"message": "Run project-manager-agent skill and report status of all sub-agents.",
"model": "openrouter/google/gemini-2.5-flash",
"thinking": "low",
"timeoutSeconds": 60
},
"schedule": {
"kind": "every",
"everyMs": 300000
},
"delivery": {
"mode": "announce"
},
"sessionTarget": "isolated",
"name": "Project Manager (Sub-agent Monitor)"
}
version: 0.2.0