The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements an adaptive learning system that ingests untrusted data from task contexts and outcomes to optimize future strategies. * Ingestion points: Data is ingested via rb.recordExperience, rb.learnPattern, and rb.recommendStrategy using task.context and task.type as seen in SKILL.md. * Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the documentation examples. * Capability inventory: The AdaptiveAgent class uses learned strategies to guide task execution, creating a loop where untrusted data influences subsequent agent actions. * Sanitization: No sanitization or validation of the ingested context or pattern data is demonstrated in the provided integration snippets.

ReasoningBank Intelligence