literature-sweep

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches paper metadata and abstracts from official CrossRef and OpenAlex API endpoints. This is standard functionality for academic literature discovery.
  • [COMMAND_EXECUTION]: Executes a local verification script verify_citation.py within the plugin directory to perform citation validation against external metadata.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external academic sources which could potentially contain malicious instructions intended to influence agent summarization.
  • Ingestion points: External paper titles and abstracts are fetched from CrossRef and OpenAlex APIs in phases/02-search.md.
  • Boundary markers: The task brief in phases/03-parallel-read.md uses structural batch markers like --- candidate <i> of <K> --- to separate individual candidate summaries.
  • Capability inventory: The skill possesses the ability to write to project files (references.bib, references.csv) and execute a local shell command for citation verification.
  • Sanitization: The engine implements basic sanitization by removing LaTeX artifacts and JATS-XML tags, and truncating abstracts before they are passed to the curator agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 09:21 AM
Security Audit — agent-trust-hub — literature-sweep