writing-beats

Pass

Audited by Gen Agent Trust Hub on Aug 11, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is designed to ingest and process 'raw material' provided by a user to generate articles. This creates an indirect prompt injection surface where malicious instructions embedded in the raw material could potentially influence the agent's behavior.
  • Ingestion points: The agent reads user-supplied markdown files ('raw material') and iteratively re-reads the 'article file' from disk (SKILL.md).
  • Boundary markers: The instructions lack explicit boundary markers or 'ignore embedded instruction' directives when processing the raw material.
  • Capability inventory: The skill has the capability to write to the local file system ('write only that beat to the article file') and read from disk.
  • Sanitization: There is no mention of sanitizing or escaping the content pulled from the raw material before it is processed or written to the article file.
  • [COMMAND_EXECUTION]: The skill instructs the agent to read from and write to the local disk. While these are standard file operations for a writing tool, they represent a capability that could be abused if coupled with malicious instructions in the input data. However, no arbitrary command execution or shell access was detected.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 11, 2026, 01:22 AM
Security Audit — agent-trust-hub — writing-beats