calibre-distill

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from Calibre books to generate executable instructions for new skills.
  • Ingestion points: Data is retrieved via calibre_get_content and calibre_semantic_search tools (SKILL.md).
  • Boundary markers: The skill includes Quality Rule #7, which instructs the agent to synthesize content rather than copying raw text, and limits verbatim quotes.
  • Capability inventory: The agent can create directories (mkdir), write files, and update the Calibre database (calibre_update_book).
  • Sanitization: The process relies on LLM synthesis; no explicit technical sanitization or escaping of the ingested book content is defined.
  • [COMMAND_EXECUTION]: The skill uses shell commands, specifically mkdir -p, to create the necessary directory structure for the generated skills on the user's local file system.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 06:00 PM
Security Audit — agent-trust-hub — calibre-distill