distill-skill

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to ingest untrusted data (bug reports, CTF writeups, blog posts) and transform it into structured instructions (SKILL.md files) for the AI agent.
  • Ingestion points: Processes arbitrary text content pasted by the user or sourced from external security reports.
  • Boundary markers: The instructions do not define strict delimiters or provide guidance to the agent to disregard instructions potentially embedded within the raw source content.
  • Capability inventory: Although the skill itself generates text, the output is intended to be saved as a new skill file which, when activated, could direct the agent to perform file operations, command execution, or network requests.
  • Sanitization: The skill contains 'Anonymization Rules' to strip PII (IPs, domains, names), but lacks mechanisms to detect or sanitize malicious natural language instructions designed to subvert the agent's behavior during the 'distillation' process.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 11:59 AM