The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for retrieving data from an external Redis store (e.g., caching, session management, and streams) and incorporating it into application logic. This introduces a vulnerability surface where data stored in Redis could be manipulated to influence the agent's behavior if later processed as a prompt.
Ingestion points: Data is ingested through commands like redis.get, redis.hgetall, redis.xreadgroup, and redis.xpending in files such as examples/caching.md, examples/data-structures.md, and examples/queues.md.
Boundary markers: The provided code examples do not include delimiters or 'ignore embedded instructions' warnings for data retrieved from Redis.
Capability inventory: The skill facilitates interaction with a Redis server and suggests patterns for using retrieved data in application logic, which could include downstream file or network operations.
Sanitization: While JSON parsing is used in the examples, there is no demonstration of content sanitization or validation of the data retrieved from external storage before it enters the application context.
Mitigation: To secure these patterns, developers should wrap external content in clear delimiters with explicit 'ignore embedded instructions' warnings and perform strict schema validation on retrieved JSON data.

api-database-redis