Answering Research Questions

Warn

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands (mkdir -p) to create local directory structures for organizing research papers and metadata.\n- [DATA_EXFILTRATION]: The skill prompts the user for an email address to be used with the Unpaywall API. While this is a requirement of the service provider, it constitutes the collection and potential transmission of PII.\n- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to generate and execute Python scripts (e.g., generate_summary.py) and code snippets to process, filter, and summarize data retrieved from external research databases.\n- [PROMPT_INJECTION]: There is a vulnerability surface for indirect prompt injection because the skill ingests untrusted data from external research papers and incorporates it into generated reports and scripts without explicit sanitization.\n
  • Ingestion points: Abstracts and full-text content from PubMed, Semantic Scholar, and Unpaywall via integrated skills.\n
  • Boundary markers: None specified in the instructions to delimit external content from internal instructions.\n
  • Capability inventory: File system writes, shell command execution, and script generation/execution.\n
  • Sanitization: No evidence of escaping or validating external text before processing or inclusion in reports.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 13, 2026, 02:28 AM