task-assessor
PassAudited by ClawScan on May 8, 2026.
Overview
This is an instruction-only task-estimation skill with no code, credentials, or install steps; the main thing to notice is its rule for when the agent should ask for confirmation.
This skill appears safe to install as an instruction-only assessment helper. Before relying on it, remember that its L1/L2 “execute directly” rule should not override normal confirmations for actions with real-world side effects, such as sending messages, modifying files, using paid APIs, or changing account data.
Findings (1)
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.
Tasks classified as simple or ordinary may proceed under the agent’s normal permissions without an extra assessment confirmation, while more complex tasks should trigger a report and confirmation request.
The skill sets a difficulty-based stopping and confirmation policy for the agent. This is disclosed and purpose-aligned, but it can influence whether the agent proceeds immediately or asks the user first.
L1 或 L2 → 直接执行，不需要确认 ... L3、L4 或 L5 → 先输出评估报告 ... 是否继续执行？
Treat the difficulty estimate as advisory and keep normal approval safeguards for actions that change files, accounts, messages, payments, or public content, regardless of the assigned L-level.