The Agent Skills Directory

[SAFE]: The skill maintains all state locally in the ~/coding/ directory and explicitly limits its scope to avoid network requests or unauthorized file access.
[COMMAND_EXECUTION]: Includes a setup instruction in SKILL.md to create a storage directory using mkdir -p ~/coding. This is a benign and necessary operation for the skill's local storage functionality.
[PROMPT_INJECTION]: The skill implements a persistent memory system that ingests user corrections to influence future responses, creating an Indirect Prompt Injection surface.
Ingestion points: Untrusted data enters the context via user corrections and confirmations as specified in SKILL.md.
Boundary markers: No specific formatting boundaries are defined for the storage of strings in memory.md beyond the ultra-compact format rules.
Capability inventory: Operational capabilities are restricted to basic file system setup (mkdir) and file operations within the ~/coding/ directory.
Sanitization: The skill does not specify sanitization or validation logic for the user-supplied preference strings.
Context: Because this behavior is the primary intended purpose of the skill and requires explicit human confirmation before storage, the risk is minimal.

Coding