Skillv1.0.0

ClawScan security

保险保单分析助手 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 9, 2026, 4:26 PM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's requested inputs and runtime instructions are coherent with an insurance‑policy analysis helper; it asks for user and policy details (including photos) which are proportionate to the stated purpose, but those inputs are sensitive so users should be cautious about what they share.
Guidance: This skill behaves like an insurance analysis assistant and will ask you to share personal details and policy images (age, health, income, debts, scanned contracts). That is necessary for meaningful analysis but is sensitive data: only share what’s strictly needed. Before using it, consider: (1) redact identity numbers, bank account numbers, or other unnecessary personal identifiers from photos; (2) avoid sending full ID documents or signed contracts unless you trust the service; (3) ask how long the data will be retained and whether it will be stored or forwarded externally; (4) prefer anonymized/test data if you only want to try the skill; and (5) for binding legal/financial decisions consult a licensed insurance professional — this tool provides guidance, not legally binding advice.

Review Dimensions

Purpose & Capability: okName/description (policy analysis) match the instructions and reference docs. All required artifacts (questionnaire, analysis guide, terms) align with performing an insurance coverage and premium evaluation. No unrelated credentials, binaries, or system access are requested.
Instruction Scope: noteInstructions explicitly ask the agent to collect personal data and accept user-sent policy images/text and to "extract key policy information." That behavior is expected for the purpose, but it involves sensitive PII and health information. The SKILL.md does not specify retention/deletion rules or how extracted images/text are handled beyond a single-line privacy statement, which is an operational/privacy gap (not a technical contradiction).
Install Mechanism: okInstruction-only skill with no install spec and no code files — lowest install risk. Nothing is downloaded or written to disk by the skill bundle itself.
Credentials: noteThe skill requests no environment variables or external credentials (proportional). However, it collects sensitive user-supplied data (age, health status, income, debts, scanned policy documents). While appropriate for insurance analysis, collecting this data increases privacy risk and requires user consent and careful handling.
Persistence & Privilege: okalways is false and the skill is user-invocable; it does not request persistent system privileges or modify other skills. Autonomous invocation remains possible per platform defaults but is not flagged on its own.