The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands, specifically the rfd4-generate tool and several setup/validation scripts like preflight.sh. These are standard operations for its stated purpose of protein design.
Evidence: SKILL.md and multiple reference files contain examples of bash and rfd4-generate command execution.
[EXTERNAL_DOWNLOADS]: The skill references a setup process that involves downloading model checkpoints and example data from a public bucket associated with the project.
Evidence: references/overrides.md mentions the rfd4-proteina-public-e04a bucket used by the download_example_data.sh script.
[INDIRECT_PROMPT_INJECTION]: The skill processes annotated CIF files and benchmark data as input. If these files are provided by untrusted third parties, they could potentially contain malicious metadata, though this is an inherent surface for scientific modeling tools.
Evidence: SKILL.md (Step 2) and references/outputs.md describe the ingestion and processing of external CIF data.

rfd4-proteina-generate