openrouter-caching-strategy
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate code examples for implementing caching strategies with the OpenRouter API. It follows best practices by using environment variables for API keys and targeting official API endpoints.
- [DATA_EXFILTRATION]: The skill facilitates network communication with OpenRouter (openrouter.ai) and a local Redis instance (localhost:6379), which is consistent with its stated purpose of LLM response caching.
- [INDIRECT_PROMPT_INJECTION]: The skill defines functions that ingest user-supplied data (prompts and message lists) which are then processed for caching and sent to the LLM provider.
- Ingestion points: Prompt and message arguments in functions such as
cached_completion,chat_with_cache, andtracked_cached_chatacross multiple files. - Boundary markers: Absent. The skill passes user input directly to the API client.
- Capability inventory: Network operations to the OpenRouter API and local Redis instance.
- Sanitization: Absent. Input strings are used directly for SHA-256 hashing and API transmission.
Audit Metadata