prompt-caching
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The implementation of Cache Augmented Generation (CAG) involves ingesting external document content and placing it directly into the LLM system prompt.
- Ingestion points: The
CAGSystem.buildCachedContextfunction inSKILL.mdaccepts and processes an array of documents for insertion into the prompt. - Boundary markers: The code uses simple Markdown headers (
## title) as delimiters but does not include explicit instructions to the model to ignore commands embedded within the document content. - Capability inventory: The skill utilizes network capabilities to interact with the Anthropic API and Redis caching layers.
- Sanitization: There is no evidence of content sanitization or escaping in the provided code snippets to mitigate the risk of embedded malicious instructions.
- The skill's author explicitly acknowledges this risk in the limitations section, advising that cached documents be treated as untrusted data.
Audit Metadata