find-literature

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill includes instructions to install the 'asta-plugins' utility directly from the official AllenAI GitHub repository using the 'uv' tool manager. \n- [EXTERNAL_DOWNLOADS]: Installation procedures and search operations involve fetching data from the vendor's GitHub repositories and the Semantic Scholar academic database. \n- [COMMAND_EXECUTION]: The skill utilizes shell commands for file system management (mkdir, mv) and uses the 'jq' utility to parse and filter JSON-formatted literature search results. \n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes and presents untrusted text from external academic publications. \n
  • Ingestion points: Paper abstracts, body snippets, and citation contexts retrieved via external search tools in SKILL.md. \n
  • Boundary markers: None identified; retrieved content is interpolated into the session context without explicit delimiters. \n
  • Capability inventory: Shell access for file manipulation (mkdir, mv) and data processing via 'asta' and 'jq'. \n
  • Sanitization: No sanitization or validation of retrieved academic content is performed prior to processing.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 12:09 AM