ingesting-academic-content

Pass

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection by processing untrusted data from external sources.\n
  • Ingestion points: Untrusted data enters the agent context through scripts/ingest-pdf.js (processing local files) and scripts/ingest-web.js (fetching content from external URLs).\n
  • Boundary markers: Absent. The extracted content is structured as JSON, but the raw text fields (e.g., fullText in scripts/ingest-pdf.js and textContent in scripts/ingest-web.js) lack delimiters or explicit warnings to the agent to disregard instructions within the data.\n
  • Capability inventory: Across all scripts, the skill's capabilities are limited to reading/writing files and performing network requests for content fetching (scripts/ingest-web.js). No subprocess spawning (exec), dynamic code execution (eval), or high-risk file system operations were detected.\n
  • Sanitization: Absent. No filtering or escaping logic is present in the scripts to detect or mitigate potential prompt injection attempts within the processed text.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 4, 2026, 01:44 PM
Security Audit — agent-trust-hub — ingesting-academic-content