The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads project source code from a well-known public repository on GitHub (github.com/karpathy/jobs) and uses Playwright to install browser binaries.
[COMMAND_EXECUTION]: The instructions involve executing a data processing pipeline consisting of local Python scripts (scrape.py, process.py, score.py, etc.) and using the uv tool for dependency management.
[CREDENTIALS_UNSAFE]: The skill recommends using a .env file to manage the OpenRouter API key. This is a secure and standard practice for development environments, ensuring sensitive credentials are not hardcoded in the source code.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external source (Bureau of Labor Statistics) which is eventually fed into an LLM for scoring.
Ingestion points: Occupation descriptions are scraped from bls.gov and stored as Markdown in pages/.
Boundary markers: No specific boundary delimiters are defined in the provided instruction snippets to separate external data from the system prompt.
Capability inventory: The pipeline performs network requests to an LLM API (score.py) and writes data to the local file system (build_site_data.py).
Sanitization: External data is processed from raw HTML to Markdown without explicit sanitization logic mentioned in the documentation, representing a potential but contextually low-risk surface.

karpathy-jobs-bls-visualizer