peer-review
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is focused on academic role-playing for critique and revision. It does not invoke any external URLs, download dependencies, or execute system-level commands.
- [PROMPT_INJECTION]: The skill instructions create a surface for indirect prompt injection by processing untrusted user papers. 1. Ingestion points: User-supplied draft text (SKILL.md). 2. Boundary markers: Absent; user text is processed without explicit structural isolation. 3. Capability inventory: Limited to text generation and saving drafts as files; no shell, network, or code execution capabilities are defined. 4. Sanitization: Absent. The threat level is minimal due to the skill's restricted capabilities.
Audit Metadata