Supervising Agents

Use when YOU are dispatching tasks to subagents (Agent tool, openclaw, parallel workers). You become the supervisor by default. This skill guides how to moni...

This skill appears coherent and focused on supervising subagents. Before installing or using it: (1) confirm the agent platform's permissions for file system access and messaging so the skill's ls/wc/find and SendMessage probes only run where you expect; (2) avoid pointing checks at sensitive directories or broad globs; (3) tune check intervals and timeouts to avoid noisy/probing behavior that could trigger rate limits or accidental data exposure; (4) be aware README suggests copying SKILL.md into ~/.claude/skills or setting CLAW_SKILLS — those modify your user environment, so review before applying; (5) prefer human escalation for tasks involving sensitive data or credentials. If you want stronger assurance, provide runtime logs or a minimal example run so you can confirm the skill's behaviors in a safe environment.

Type: OpenClaw Skill Name: supervising-agents Version: 1.0.0 The 'supervising-agents' skill provides a structured framework for an AI agent to manage and verify the work of subagents. It defines a 5-phase workflow (Prepare, Dispatch, Monitor, Intervene, Verify, Report) and uses standard system utilities like 'ls', 'wc', and 'grep' (in SKILL.md) to monitor file-based progress and validate outputs. The instructions focus on ensuring task accountability and preventing 'lazy' AI behaviors without any indicators of data exfiltration, unauthorized execution, or malicious intent.