The Agent Skills Directory

[COMMAND_EXECUTION]: The script _status_core.py uses subprocess.run to communicate with the platform's script executor. This mechanism is used for inter-skill operations such as session binding and updating terminal titles. The implementation uses a structured command list without a shell, which is a secure practice that prevents command injection. The parameters passed to the subprocess are either hardcoded constants or validated identifiers.
[PROMPT_INJECTION]: The skill ingests data from request.md to perform deterministic classification and routing. This ingestion is handled via keyword scoring and regular expression matching rather than direct prompt interpolation into an LLM. As the processing logic is algorithmic and does not follow instructions embedded in the data, it is not susceptible to indirect prompt injection.
[SAFE]: The skill implements various file operations, including directory deletion (delete-plan) and archiving, which are strictly scoped to the .plan directory. It includes protective measures such as kebab-case validation for plan identifiers and path-traversal checks when moving files during lesson restoration.

manage-status