Skillv0.1.0

ClawScan security

叶武滨分身 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 16, 2026, 1:41 AM

Verdict: benign
Confidence: medium
Model: gpt-5-mini
Summary: This is an instruction-only persona skill that is internally consistent with its stated purpose (time-management / coaching roleplay) and requests no credentials or installs, but it instructs the agent to impersonate a named real-world figure and to promote course offerings, which you should be aware of before installing.
Guidance: This skill is low-risk from an installation/credential perspective because it's instruction-only and requests nothing sensitive. Before installing, consider these practical issues: 1) Impersonation and authorization — the skill directs the agent to speak as the named person using first-person phrasing; confirm you have permission to represent that individual or that users are clearly informed it's a persona. 2) Promotional behavior — the skill encourages guiding users to the provider's courses; decide whether you want an installed skill that upsells or funnels users to external offerings and ensure disclosures are present. 3) Veracity of claims — the SKILL.md references specific documents, case studies, and audience numbers; verify sources or add guardrails to avoid hallucinated factual claims. 4) Test before wide use — run sample interactions to ensure the agent stays within declared boundaries (exits role on command, avoids medical/financial advice as stated). If you need higher assurance, request evidence of authorization to impersonate the named individual and require the skill to include explicit user-facing disclosure that responses are roleplay-based and not direct quotes from the person.

Review Dimensions

Purpose & Capability: okName/description claim (a coach/persona for time management) matches the SKILL.md instructions: role, workflows, heuristics and stylistic rules. No binaries, env vars, or installs are requested — all proportionate for a persona/instruction-only skill.
Instruction Scope: noteThe SKILL.md gives detailed runtime instructions for roleplay (use “I”, specific sentence templates, behavioural rules, stepwise workflow) which are appropriate for producing a stylistic persona. Notable concerns: it explicitly directs the agent to present as a named person (叶武滨) and to 'naturally' guide users toward the provider's course system—this is promotional and may amount to impersonation if the persona is a real individual or uses trademarked material without authorization. The instructions also require the agent to cite internal case examples and specific course material claims (e.g., attendance numbers) that could lead to hallucinated or unverifiable assertions if sources aren't available.
Install Mechanism: okNo install spec, no code files, no downloads. Lowest-risk install profile — nothing will be written to disk or fetched at install time.
Credentials: okNo environment variables, credentials, or config paths required. The skill does not request access to unrelated services or secrets.
Persistence & Privilege: okalways is false and the skill does not request elevated persistence or modification of other skills or agent configurations. Autonomous invocation is allowed (platform default) but not combined with other privileges.