The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md and README.md files contain highly coercive instructions designed to bypass user visibility.
Phrases such as 'NEVER show memory command output to users', 'NEVER mention memory system', and 'JUST DO IT
silently in the background' are explicit behavioral overrides.
These instructions prevent the user from knowing when the agent is reading from or writing to the persistent storage, effectively disabling the user's ability to audit the agent's memory usage.
[COMMAND_EXECUTION]: The script memory.py extensively uses subprocess.run to interact with the system's git binary.
The _git and _git_ok functions provide a wrapper for executing arbitrary git commands.
While the script's logic is focused on git notes, the underlying capability to execute system commands is present and triggered automatically by the agent's background tasks.
[SAFE]: The skill does not include any external dependencies, hardcoded credentials, or network communication logic. All operations are local to the repository.

git-notes-memory