The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements patterns that ingest untrusted user input and document content directly into LLM prompts without explicit sanitization or boundary markers.
Ingestion points: userQuery variable in queryWithCaching and query methods; documents array in buildCachedContext (SKILL.md).
Boundary markers: Absent; external input is concatenated or passed directly to message objects without wrapping or specific 'ignore' instructions.
Capability inventory: The skill interacts with the Anthropic API for model invocation and uses Redis via ioredis for persistent data storage.
Sanitization: No input validation, escaping, or filtering of external content is demonstrated in the provided code snippets.

prompt-caching