orchestrator
PassAudited by VirusTotal on May 4, 2026.
Overview
Type: OpenClaw Skill Name: ah-orchestrator Version: 1.0.0 The skill bundle defines an 'orchestrator' agent designed to coordinate complex tasks by routing them to specialized sub-agents (e.g., /backend-architect, /security-auditor). The instructions in SKILL.md and references/examples.md focus on project management, parallel execution strategies, and human-in-the-loop checkpoints, with no evidence of malicious intent, data exfiltration, or unauthorized command execution.
Findings (0)
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.
A task may be handed to different specialist agents than the user initially expected.
The skill can route work through fallback agents, which is core to an orchestrator but means task authority may broaden if a primary agent fails.
Primary Agent (95%+) ↓ if unavailable or fails Secondary Agent (70%+) ↓ if unavailable or fails Generalist Fallback (/fullstack-engineer) ↓ if still fails Multi-Agent Coordinator (/multi-agent-coordinator)
Review the proposed agent plan and fallback chain before allowing downstream agents to make changes.
Outputs from one agent session could influence another phase of work, including mistakes or unsafe assumptions.
The skill encourages parallel work across multiple agent sessions and later result aggregation. This is consistent with its purpose, but cross-session context and outputs should be reviewed before being combined.
For User: Open multiple Claude Code sessions to run these in parallel: Session 1: /backend-architect Design the API; Session 2: /ux-designer Create user flows; Session 3: /data-engineer Plan data pipeline
Treat each agent’s output as untrusted until reviewed, and avoid sharing secrets or unnecessary private project data across parallel sessions.
Stale or sensitive checkpoint content could be reused later if the agent stores or summarizes project state.
Checkpointing can create persistent summaries or saved state. The behavior is disclosed and purpose-aligned, but the visible artifacts do not specify storage location, retention, or reuse boundaries.
Automatic Checkpoints (V4) Auto-save progress at key milestones for disaster recovery.
Confirm where checkpoints are saved, avoid including secrets in checkpoints, and periodically review or clear saved project context.
Users may overestimate the reliability or safety of the orchestration plans.
The description uses strong authority and quality claims that are not substantiated by the provided artifacts.
powered by proven agentic design patterns from 1K+ real-world AI projects, enhanced with industry-leading multi-agent coordination
Treat generated plans as recommendations, not verified guarantees, and require explicit review before implementation.