The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a significant attack surface for Indirect Prompt Injection by processing untrusted data.
Ingestion points: The skill ingests untrusted data through the learn_from_session and adapt_responses methods, which analyze interaction history and user feedback.
Boundary markers: No delimiters or explicit instructions to ignore embedded commands are present in the logic for processing session data.
Capability inventory: The skill has access to high-privilege tools including Bash, Write, Glob, Grep, and Read.
Sanitization: There is no evidence of input validation or escaping of external content before it is used to influence system logic.
[COMMAND_EXECUTION]: The skill utilizes dynamic execution patterns to modify its own behavior and the agent environment.
Execution method: Functions like implement_evolution_step, deploy_knowledge_updates, and implement_optimization suggest the runtime application of logic or configurations derived from patterns learned from user data.
Risk factor: The combination of runtime-determined behavior with the Bash and Write tools creates a pathway for arbitrary command execution and unauthorized system modification if learning signals are subverted.

moai-learning-optimizer