Delegate

PassAudited by ClawScan on May 1, 2026.

Overview

This instruction-only skill is a coherent delegation guide; its main risk is that delegated sub-agents may receive scoped context and, for user-requested code tasks, may modify files.

This skill appears safe to install as an instruction-only delegation helper. Before using it, remember that sub-agents may see the context you provide and may edit files for delegated coding tasks, so keep shared context minimal and review any file changes before accepting them.

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.

What this means

Task details, files, or research data may be handled by a sub-agent rather than only the main agent.

Why it was flagged

The skill explicitly routes work to sub-agents and passes task context to them. The instruction to avoid full history is a good boundary, but users should understand that selected files or information may be shared with delegated agents/models.

Skill content
Every spawn must include: ... CONTEXT: Only files/info needed (never full history)
Recommendation

Delegate only the minimum necessary context, avoid including secrets, and review which files or information are provided to sub-agents.

What this means

If used on a real project, a delegated agent may change repository files and commit them as part of the requested task.

Why it was flagged

A code implementation template allows a delegated sub-agent to modify files and create a commit. This is purpose-aligned for delegated coding tasks, but it is a mutable action that should remain user-directed and reviewable.

Skill content
OUTPUT: Modified files, tests passing
DONE: Commit with message "[feature] implemented"
Recommendation

Review diffs and test results before accepting delegated code changes, and require explicit user approval before commits in important repositories.