autoresearch

Pass

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local script provided by the author, ./scripts/boundary-score.py, to identify and rank frontier pages for topic selection.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it uses the WebFetch tool to ingest untrusted information from external websites and incorporates this data into its synthesis without sanitization.
  • Ingestion points: External web content retrieved during the iterative rounds of the research loop.
  • Boundary markers: Absent. The instructions do not define delimiters or provide warnings to ignore malicious instructions within the fetched data.
  • Capability inventory: The skill utilizes Read, Write, Edit, WebSearch, and WebFetch permissions to manage the wiki and access the internet.
  • Sanitization: Absent. There is no mention of filtering, validation, or escaping the retrieved content before it is used to generate or update wiki pages.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 25, 2026, 08:58 AM