The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script to retrieve session statistics for the 'undo' feature.
Evidence: The skill constructs a path using environment variables (e.g., CLAUDE_PLUGIN_ROOT) and executes python3 "$SCRIPT_DIR/session_stats.py" peek.
Context: This is a local administrative operation intended to track recent memory IDs for the session, allowing the user to undo recent additions.
[SAFE]: The skill implements a safety-first approach to data management by requiring explicit user confirmation before any memory is deleted.
Evidence: 'Never delete without confirmation. This is destructive.' instructions in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it retrieves and displays stored memories which could contain malicious content.
Ingestion points: Memory content is fetched via get_memory and search_memories in SKILL.md.
Boundary markers: There are no explicit delimiters used when displaying memory content to the user.
Capability inventory: The skill can perform deletions and execute local scripts.
Sanitization: No specific sanitization of retrieved memory content is mentioned before it is displayed in the agent's context.

forget