The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes the official @google/genai Node.js package to interact with Google's Gemini API infrastructure. These external resources are provided by a trusted organization and are considered safe for use.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external documents (PDF, Word, Code) for semantic search. 1. Ingestion points: File uploads via uploadToFileSearchStore in SKILL.md and scripts/create-store.ts. 2. Boundary markers: No specific delimiters are required in the prompt instructions to isolate document content. 3. Capability inventory: The agent performs generateContent calls using the fileSearch tool to query the indexed data. 4. Sanitization: Document processing and safety filtering are handled by the service provider's internal API. As this represents the primary and intended functionality of the RAG skill, the severity is assessed as safe.
[SAFE]: No malicious patterns, obfuscation, or unauthorized privilege escalation attempts were detected. The CLI scripts provided for store management follow standard software engineering practices.

google-gemini-file-search