Create SubAgent

PassAudited by ClawScan on May 10, 2026.

Overview

This instruction-only skill appears aligned with creating and managing SubAgents, but users should notice that it can spawn tool-using or persistent subagents and the README references a script that is not included.

Install this only if you intentionally want your agent to create and manage SubAgents. Use narrow task descriptions, prefer one-off run mode unless you need persistence, monitor logs, kill SubAgents when finished, and do not run the README’s referenced PowerShell script unless you obtain and verify it separately.

Findings (4)

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 spawned SubAgent could read or write files or run Python code as part of the assigned task.

Why it was flagged

The data-analysis SubAgent template explicitly names execution and file tools. This is expected for a data-analysis/development helper, but it means spawned agents may perform impactful actions if those tools are available.

Skill content
工具:exec (Python), read, write
Recommendation

Give SubAgents narrow tasks, specify allowed files or directories, and review before allowing code execution or file modification.

What this means

Information you send to the SubAgent may appear in SubAgent context or logs.

Why it was flagged

The skill creates separate SubAgent sessions and provides commands for sending messages and viewing logs. This is central to the skill, but it creates inter-agent communication paths where user data or task context may be shared.

Skill content
调用 sessions_spawn ... /subagents send <id|#> "消息内容"
Recommendation

Avoid sending secrets or sensitive files unless needed, and review SubAgent logs and messages when handling private data.

What this means

A SubAgent session or its retained artifacts may outlive the initial request until it is archived or killed.

Why it was flagged

The skill supports persistent SubAgent sessions and keeps cleanup artifacts by default. It also discloses kill commands and automatic archiving, so this is not hidden behavior, but persistence should be understood.

Skill content
`session` - 持久会话 ... cleanup: "keep"
Recommendation

Prefer run mode for one-off work, use `/subagents kill` when finished, and choose cleanup settings that match your privacy needs.

What this means

If you find or download a similarly named script elsewhere, it was not part of the reviewed artifacts.

Why it was flagged

The README documents a PowerShell helper script, but the supplied manifest contains no such script and the package is described as instruction-only. This is a documentation/provenance gap, not evidence of malicious code.

Skill content
.\create-subagent.ps1 -preset dev
Recommendation

Do not run any external create-subagent script unless it is obtained from a trusted source and reviewed separately.