The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple Python scripts (local_artifact_smoke.py, notebooklm_artifact_matrix.py, notebooklm_health.py) utilize subprocess.run to execute external binaries, including node, notebooklm CLI, and system utilities like pgrep for environment health checks. These are used with structured argument lists rather than raw shell strings.
[COMMAND_EXECUTION]: The script local_artifact_smoke.py dynamically constructs and executes a Node.js script using the node -e flag to automate PDF and image generation via the Playwright library.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies including the notebooklm CLI tool and the playwright Node.js package to generate and render artifacts.
[PROMPT_INJECTION]: The skill ingests article content from external files or URLs, creating an indirect prompt injection surface when this content is interpolated into LLM prompts for generating quizzes and summaries.
Ingestion points: Article content is read from local Markdown files or fetched from URLs as specified in SKILL.md and local_artifact_smoke.py.
Boundary markers: No explicit safety delimiters are used when interpolating article excerpts into the generated reports or quiz prompts.
Capability inventory: The skill possesses local file system write access, execution of shell commands via subprocess, and network interaction through the notebooklm CLI.
Sanitization: The skill uses html.escape to sanitize article content before rendering HTML artifacts and json.dumps to safely inject paths into dynamically generated JavaScript snippets.

soia-pkm-transform-article-notebooklm