The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements an interactive browser session that records user events, creating a potential surface for indirect prompt injection.\n
Ingestion points: The agent is directed by references/visual-session.md to read user interaction data from the $STATE_DIR/events file.\n
Boundary markers: There are no explicit instructions to use delimiters or boundary markers when the agent interpolates data from the events log into its prompt.\n
Capability inventory: The skill includes scripts for starting and stopping a Node.js server (scripts/start-server.sh, scripts/stop-server.sh), writing HTML files to a local directory, and executing Mermaid diagram rendering scripts (scripts/render_mermaid.py).\n
Sanitization: While the scripts/server.cjs server uses standard JSON serialization for events, it does not perform content-specific sanitization or filtering of the user-provided data before it is logged for the agent's use.\n- [EXTERNAL_DOWNLOADS]: The Python utilities render_mermaid.py and validate_mermaid.py utilize the uv tool to dynamically fetch and install the mermaidx package from the Python Package Index (PyPI). PyPI is a standard and well-known package registry.\n- [COMMAND_EXECUTION]: The skill uses local shell and Python scripts to manage its background server process and perform diagram validation and rendering tasks as part of its intended functionality.\n- [DYNAMIC_EXECUTION]: The skill generates HTML fragments or documents from internal templates to provide visual mockups, which are then rendered in a local browser session. This represents standard dynamic content generation for the skill's primary visualization purpose.

show-me