writing-beats
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill documentation uses the term 'exploit' to describe a specific phase of the writing process (mining raw material for content). Analysis confirms this is a stylistic choice related to the 'explore/exploit' paradigm and not a reference to security exploits or malicious behavior.\n- [PROMPT_INJECTION]: The skill processes untrusted 'raw material' provided by the user, which presents a surface for indirect prompt injection. However, this is inherent to the skill's purpose. The risk is effectively mitigated by the human-in-the-loop design where the agent only writes to the file system after the user explicitly selects a 'beat' from several candidates.\n
- Ingestion points: User-supplied markdown file containing raw material.\n
- Boundary markers: None present for the raw material input.\n
- Capability inventory: Reading from and writing to local files as directed by the user.\n
- Sanitization: The workflow relies on user selection and review as a manual filter for content.
Audit Metadata