writing-beats

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted content from 'raw material' files and re-reads its own output from disk to determine subsequent actions. It lacks instructions for the agent to ignore potentially malicious commands embedded within that data.
  • Ingestion points: Reads a markdown file of raw material (SKILL.md) and re-reads the article file from disk before every write (SKILL.md).
  • Boundary markers: None present; the agent is instructed to 'mine the pile' and 'pull material' without delimiters or safety warnings regarding the content.
  • Capability inventory: Perform file-read and file-write operations to persistent storage (SKILL.md).
  • Sanitization: No validation or sanitization is performed on the ingested markdown material.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 10:27 AM
Security Audit — agent-trust-hub — writing-beats