Coordinate human-in-the-loop agent teams in auditable Matrix rooms with HiClaw
Run manager-worker agent collaboration in Matrix rooms where humans can watch, intervene, and keep credentials out of worker hands.
npx skills add agentskillexchange/skills --skill coordinate-human-in-the-loop-agent-teams-in-auditable-matrix-rooms-with-hiclaw
HiClaw is a manager-worker collaboration system for running multiple agents inside Matrix rooms with full human visibility and intervention. Use this when a task needs auditable handoffs, live supervision, and secure worker isolation across several agents, not when you just want to chat with one agent or adopt a general-purpose SDK. The scope boundary here is specific: this skill is about coordinating multi-agent work through transparent Matrix-room workflows and human checkpoints, rather than listing HiClaw as a broad agent platform.
What this skill actually does
What the agent actually does:
– Creates a Manager/Worker collaboration pattern for complex tasks
– Routes agent work through Matrix rooms that humans can inspect in real time
– Keeps real credentials behind the gateway so workers do not hold API keys or PATs
– Lets operators step into rooms, correct behavior, and supervise execution
When to invoke it:
– You need several agents to collaborate on one task with visible handoffs
– You need human-in-the-loop review without losing chat history or coordination context
– You want room-based collaboration and secure credential pass-through instead of opaque agent-to-agent orchestration
Why this is skill-shaped instead of a product card:
– The publishable unit is the auditable Matrix-room coordination workflow
– The user-facing job is operating supervised multi-agent task rooms, not ‘install a platform’
– The upstream docs repeatedly center Matrix, manager-worker decomposition, human supervision, and credential isolation as the invocation shape