ACP Orchestrator

PassAudited by ClawScan on May 1, 2026.

Overview

The skill is a clearly disclosed AI-agent orchestrator, but users should review its broad parallel/subagent execution and external CLI package dependencies before use.

This skill appears coherent and purpose-aligned for orchestrating AI agents from the terminal. Before installing, verify the external `acpx` and optional agent CLI packages, review any task files, and be careful with parallel/subagent workflows because they may delegate your code or prompts to multiple agents.

Findings (3)

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.

What this means

A task file or passthrough command could cause several installed agents to act on the user's workspace at once.

Why it was flagged

The skill intentionally exposes broad orchestration and passthrough commands that can run tasks through other agents, including from a file. This is aligned with the orchestrator purpose, but users should review inputs before running them.

Skill content
`parallel` | Run agents parallel from file | `acpx parallel tasks.txt` ... `exec` | Direct acpx passthrough | `acpx exec opencode "task"`
Recommendation

Use this skill only with task files and agent commands you have reviewed, and prefer scoped workflows when possible.

What this means

The skill's real behavior depends on externally installed `acpx` and agent CLIs that are not included in the reviewed artifacts.

Why it was flagged

The manifest depends on external CLI packages and agent tools, while the registry metadata says there is no install spec and no required binary. This is a setup/provenance gap to notice, not evidence of malicious behavior.

Skill content
"requirements": { "binaries": ["acpx"], "optional": ["opencode", "pi", "kimi", "kilo", "codex", "claude", "gemini"] }, "install": { "npm": ["acpx"]
Recommendation

Verify the npm package and any optional agent CLIs from trusted sources before installing or using them.

What this means

Prompts, code context, or task details may be delegated to whichever installed agents are invoked.

Why it was flagged

The skill documents spawning subagents and coordinating multiple AI agents. This is expected for an orchestrator, but the artifacts do not define data boundaries between agents.

Skill content
sessions_spawn( task="acpx parallel tasks.txt", label="parallel-jobs", runtime="subagent", mode="run" )
Recommendation

Avoid including secrets in tasks, and confirm which agents will receive the work before using parallel or subagent workflows.