The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/local_artifact_smoke.py executes shell commands using subprocess.run(). Specifically, it invokes a Node.js process to run a JavaScript snippet that automates Playwright for rendering image and PDF artifacts.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to download and install several external dependencies, including playwright, notebooklm-py, and optional skills from the soia-team repository to enable specific visual providers.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted article content and interpolates it into complex prompt templates for visual generation. * Ingestion points: External article data provided via file paths or URLs is parsed and injected into prompts defined in references/prompt-*.md. * Boundary markers: The prompt templates use structural labels but lack explicit delimiters or "ignore embedded instructions" escape sequences to isolate article content from core instructions. * Capability inventory: The skill environment includes file system write access for saving generated artifacts, subprocess execution capabilities (node/playwright), and network access via the notebooklm CLI. * Sanitization: Content is minimally escaped for HTML reports, but no significant filtering or instruction-sanitization is applied to the raw article text before it is interpolated into LLM prompts.

soia-pkm-transform-article-visual