The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess calls in scripts/dashboard_setup.py and assets/dashboard/server.py to manage the local server's lifecycle. It checks port status using lsof and executes internal utility scripts (e.g., knowledge_ingest.py, record_published.py) to process data within the user's knowledge base. These operations are performed locally and are essential for the dashboard's functionality.
[INDIRECT_PROMPT_INJECTION]: The dashboard processes and renders Markdown content from the user's local vault. This creates a surface where malicious content in a note could attempt to influence the agent's behavior or manipulate the dashboard's state flow through frontmatter metadata.
Ingestion points: assets/dashboard/server.py (functions note_payload and load_reviews read and parse user-controlled Markdown files).
Boundary markers: Absent; the dashboard renders note content directly into the UI.
Capability inventory: Subprocess execution for data processing and file writing (atomic_write) to the local vault.
Sanitization: Employs contain_in_vault in assets/dashboard/server.py to prevent path traversal attacks outside the designated knowledge base directory.
[PERSISTENCE]: The skill includes documentation and XML templates for setting up launchd persistence on macOS (autostart). It explicitly states that this must be configured manually by the user and is not automated by the skill's scripts.
[DYNAMIC_EXECUTION]: assets/dashboard/bootstrap.py uses importlib.util to dynamically load server.py to extract configuration constants during the initialization of a new vault environment.

ray-dashboard