writing-beats

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided data from a raw material file and an article file on disk, which could contain malicious instructions that the agent might inadvertently execute while generating text.
  • Ingestion points: The agent is instructed to read from a markdown file of "raw material" and periodically re-read an "article file" from disk to maintain context and rhythm (SKILL.md).
  • Boundary markers: The instructions do not define delimiters or explicit warnings to the model to ignore potential instructions embedded within the raw material or the existing article text.
  • Capability inventory: The skill utilizes file system operations to read local files and write generated content back to disk ("write only that beat to the article file", "Append one beat at a time").
  • Sanitization: There is no mention of filtering, escaping, or validating the input data from the "raw pile" before it is used to populate beats in the output file.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 06:39 PM