The Agent Skills Directory

[SAFE]: The skill uses platform-native memory_read and memory_write permissions to store and retrieve user settings. There are no network operations, hardcoded credentials, or shell command executions.
[PROMPT_INJECTION]: The skill implements a preference learning mechanism that creates a surface for persistent indirect prompt injection (Category 8) by storing untrusted user strings to guide future behavior. This is handled as safe within the context of the skill's primary purpose.
Ingestion points: User input from triggers like 'I prefer', 'I like', and 'remember that I' defined in SKILL.md.
Boundary markers: None identified.
Capability inventory: memory_read and memory_write for persistence.
Sanitization: No sanitization of preference strings is specified.

preference-learner