The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs shell scripts and common CLI utilities (like find, wc, jq, and rm) to perform system diagnostics and manage local files. These commands are focused on the user's Claude Code configuration and memory storage locations.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file uses the !bash syntax to run scope-report.sh at load time. This diagnostic script is used to provide the agent with a snapshot of the current environment's memory settings and file existence without making any changes.
[DATA_EXPOSURE]: The skill interacts with configuration files such as settings.json and settings.local.json to read and update auto-memory preferences. It also enumerates the contents of the ~/.claude/projects/ directory to identify memory stores associated with various repositories.
[INDIRECT_PROMPT_INJECTION]: The skill identifies and processes markdown files generated by the agent (auto-memory) to provide status reports.
Ingestion points: Reads MEMORY.md and other topic-specific markdown files in the memory directories.
Boundary markers: No explicit boundary markers are used when analyzing these files for line counts or existence.
Capability inventory: Includes file system modification capabilities such as deletion (rm), moving files (mv), and creating backups (cp).
Sanitization: The skill uses jq for structured JSON manipulation of settings files but does not sanitize the contents of markdown files read for diagnostic purposes.

stateless