The Agent Skills Directory

[SAFE]: No security threats were identified. The skill performs standard file management operations within a local application data directory (~/.smtm/projects/).
[PROMPT_INJECTION]: The skill facilitates 'Indirect Prompt Injection' by design, as its core purpose is to store user-provided observations that influence the behavior of other skills in future sessions. This is a functional feature for persistent memory rather than a security vulnerability.
Ingestion points: The skill reads data from ~/.smtm/projects/{slug}/learnings.jsonl (SKILL.md) and accepts user input via the add command.
Boundary markers: No explicit boundary markers or isolation instructions are defined for the injected learning content.
Capability inventory: The skill instructions involve reading and writing to local files in a project-specific directory (SKILL.md).
Sanitization: The skill implements basic error handling for JSONL formatting but does not include sanitization for instructions embedded within the natural language patterns it stores.

money-learn