The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in monitor_archiver.py to trigger operations in related skills, such as episodic-archiver and skill-lab. These commands involve unsetting environment variables and setting specific working directories to manage tool execution.
[DYNAMIC_EXECUTION]: In monitor_archiver.py, the _connect function dynamically loads a Python module from a file path (MEMORY_SKILL_DIR / 'db.py') using importlib.util. This is used to share database connection logic with the memory skill but introduces a dependency on the integrity of that external file at runtime.
[COMMAND_EXECUTION]: The register-nightly command executes a shell command to register itself with a scheduler skill, passing shell-style arguments including path strings and cron expressions.
[INDIRECT_PROMPT_INJECTION]: The skill processes session transcripts and failure episodes to generate health metrics and behavioral profiles. While it uses LLM analysis for taxonomy, the potential for malicious data in processed transcripts to influence agent behavior is present, though typical for an analysis skill.

monitor-episodic-archiver