The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a structured interface for human input (text and images) which is then returned to the agent in JSON format. This represents a standard surface for indirect prompt injection where a user's response could influence the agent's subsequent logic.
Ingestion points: server.py handles POST requests from the HTML form; tui.py handles user input via Textual widgets.
Boundary markers: The skill returns data as a structured JSON object, providing clear boundaries for the agent.
Capability inventory: The skill is capable of writing session data to the local sessions/ directory via interview.py.
Sanitization: The skill performs basic HTML escaping for code blocks in server.py, though it lacks rigorous escaping for all user-provided labels in the HTML view.
[EXTERNAL_DOWNLOADS]: The HTML form template (templates/form.html) loads the Tailwind CSS library from cdn.tailwindcss.com, which is a well-known and trusted service.
[SAFE]: User responses and session metadata are stored locally in the sessions/ folder within the skill directory for recovery purposes, which is standard behavior for local agent utilities.

interview