The Agent Skills Directory

External Resource Integration: The skill fetches structural data and visualization libraries from established scientific and technology providers, including UniProt, RCSB PDB, the AlphaFold Protein Structure Database, and the Molstar CDN. These interactions are fundamental to the tool's scientific purpose.
Software Installation Workflow: The instructions guide the agent through standard environment setup using package managers like pip and npm, as well as containerization with Docker. These steps are necessary to deploy the scoring server and proxy backend.
Indirect Prompt Injection Surface: This skill processes user-supplied protein sequences and accession IDs which enter the context through the web UI and proxy backend (e.g., in App.js and app.py). While formal boundary markers are not used, the skill implements a specialized sanitization module (sequence_screening.py) to block known biological toxins. Capabilities associated with this data surface include network requests to inference endpoints and molecular analysis via mdtraj (in trajectory_analysis_real.py).
Security-Conscious Instructions: The skill explicitly instructs the agent never to ask users to paste secrets into chat and provides guidance on using placeholder credentials for local development, demonstrating good security hygiene for AI-assisted workflows.

msresearch-bioemu