The Agent Skills Directory

[DATA_EXPOSURE]: The skill manages configuration and credentials by reading from a vendor-specific configuration file located at ~/.pontistudios.studio.yml. This is a standard practice for secret management within the author's ecosystem.
[DATA_EXPOSURE]: The skill performs file operations within the user's home directory, specifically writing video drafts to ~/Documents/studio-drafts and maintaining usage logs in ~/.hominem/ai_usage.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted master scripts via the --script argument, which are then interpolated into the LLM prompt.
Ingestion points: Master script JSON files read in scripts/lib/run.mjs.
Boundary markers: Uses Markdown formatting and explicit labels ("Approved master script:") to separate script content from instructions.
Capability inventory: Local file writes to specific directories; no system command execution or arbitrary code evaluation.
Sanitization: Implements strict schema validation for both input and output using a shared validation library from the sibling studio-core project.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to OpenRouter, a well-known service, for text generation. These requests are restricted to the primary purpose of the skill.

studio-scenes