The Agent Skills Directory

[DYNAMIC_EXECUTION]: The RedisConfig.java template uses an unsafe Jackson polymorphic deserialization configuration. Specifically, it calls activateDefaultTyping with ObjectMapper.DefaultTyping.NON_FINAL and LaissezFaireSubTypeValidator.instance. This configuration is a known security risk that can lead to Remote Code Execution (RCE) if an attacker can manipulate the data stored in the Redis instance, as it permits the deserialization of arbitrary gadget classes during JSON processing.
[INDIRECT_PROMPT_INJECTION]: The skill implements patterns that ingest untrusted data into Redis operations without explicit sanitization or boundary enforcement, creating a surface for indirect prompt injection or key collision attacks.
Ingestion points: Untrusted identifiers such as identifier in RateLimiter and orderId in OrderCacheService are used to construct Redis keys.
Boundary markers: The code relies on simple string concatenation (e.g., "ratelimit:" + identifier) without ensuring the input does not contain characters that could break the key namespace.
Capability inventory: The skill uses RedisTemplate to perform opsForValue().set(), increment(), and delete() operations.
Sanitization: No sanitization or validation of the input strings is performed before they are used in key construction or value retrieval.

spring-data-redis