evidence-draft

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill performs local data transformation on research artifacts (notes, briefs, and citations) and does not possess network capabilities or external dependencies.
  • [SAFE]: The main execution logic in scripts/run.py is deterministic and avoids unsafe dynamic execution patterns like eval, exec, or direct shell command invocation.
  • [SAFE]: Robust input sanitization is performed by the _sanitize_source_text function, which applies hygiene regex patterns to remove URLs, self-narration, and boilerplate content from paper snippets.
  • [SAFE]: The skill manages potential indirect prompt injection risks by sanitizing external content before it is structured into packs for downstream use.
  • [SAFE]: The skill adheres to strict validation rules and an explicit schema for its JSONL outputs, ensuring predictable behavior within the agent environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 06:47 AM