subagent-orchestrator
PassAudited by ClawScan on May 10, 2026.
Overview
This instruction-only skill is coherent for subagent autopilot orchestration, but users should set clear task boundaries because it can keep delegating work and write persistent memory records.
Install only if you want OpenClaw to continue multi-step subagent workflows with less prompting. Keep AGENT_TASK_BOARD.md narrowly scoped, define clear stop conditions, and review MEMORY.md and daily memory records for sensitive or stale instructions.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
An active autopilot line may continue through multiple delegated steps before asking again, depending on the task board and stop conditions.
The skill intentionally enables chained subagent dispatch without asking the user to restate each step. This is purpose-aligned, but users should ensure task lines are clearly scoped.
For every autopilot line, the main controller decides the next slice and spawns it.
Use clear task-board entries, end conditions, and need_user or blocked gates for risky, destructive, public, or account-impacting work.
Incorrect, sensitive, or overly broad memory entries could persist and affect future agent decisions.
The skill writes durable task records and collaboration rules that may be reused in later sessions.
Before forgetting them, write the durable record into `memory/YYYY-MM-DD.md`... Promote stable collaboration rules into `MEMORY.md` when needed.
Review memory records periodically and avoid storing secrets, unnecessary private data, or unverified instructions in persistent memory files.