embeddings-and-search

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill follows security best practices for secret management by retrieving API keys (OPENAI_API_KEY, BRAVE_API_KEY, MISTRAL_API_KEY) from environment variables using os.environ rather than hardcoding them.
  • [INDIRECT_PROMPT_INJECTION]: The skill exposes ingestion points for external data via web search results and OCR text. Mandatory Evidence Chain:
  • Ingestion points: The client.search and client.ocr methods in SKILL.md ingest data from external providers.
  • Boundary markers: The examples do not implement delimiters or 'ignore' instructions for the retrieved data, as the examples are limited to basic metadata printing.
  • Capability inventory: The skill utilizes the liter-llm library for network-based data retrieval.
  • Sanitization: No explicit sanitization or filtering of external content is shown in these documentation snippets.
  • [SAFE]: No signs of obfuscation, remote code execution from unknown sources, persistence mechanisms, or privilege escalation were detected. The usage of the liter-llm library for embeddings and search functionality is consistent with the stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 10:22 PM
Security Audit — agent-trust-hub — embeddings-and-search