writing-beats
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content (referred to as the 'raw material' pile) to generate article beats. This creates a surface for indirect prompt injection if the material contains instructions intended to manipulate the agent's behavior.
- Ingestion points: Raw material markdown file mentioned in SKILL.md.
- Boundary markers: Absent; the agent is instructed to 'mine the pile' directly.
- Capability inventory: Reads from and writes to a local article file as specified in SKILL.md.
- Sanitization: No explicit sanitization or filtering of the raw material is defined.
- [SAFE]: The mention of 'exploit' in the skill description and instructions refers to the algorithmic concept of utilizing existing resources (exploration vs. exploitation) rather than a security vulnerability.
- [SAFE]: File system access is limited to reading and appending to the specific article file identified by the user, which is necessary for the skill's primary purpose.
Audit Metadata