The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external antibody sequences to perform analysis, structure prediction, and optimization, creating a surface for potential indirect prompt injection attacks.
Ingestion points: The agent accepts VH (Heavy) and VL (Light) chain sequences from the user in Phase 1 for annotation and subsequent engineering phases.
Boundary markers: The instructions do not specify the use of delimiters or specific warnings to ignore instructions that might be embedded within biological sequences or FASTA headers.
Capability inventory: The skill performs file operations (writing antibody_optimization_report.md and optimized_sequences.fasta) and executes complex modeling via AlphaFold_get_prediction.
Sanitization: There is no mention of validation or sanitization for input sequences to ensure they do not contain adversarial text patterns.
[EXTERNAL_DOWNLOADS]: The skill facilitates data retrieval from established scientific and biological resources to support humanization and structural analysis.
Evidence: Utilizes tools to query external databases including UniProt (protein information), IEDB (epitope data), IMGT (germline sequences), and SAbDab (antibody structures).
[COMMAND_EXECUTION]: The skill requires the agent to generate and manage multiple local files to store analysis results and candidate sequences.
Evidence: Instructions mandate the creation of a progress-updated antibody_optimization_report.md, along with optimized_sequences.fasta, humanization_comparison.csv, and developability_assessment.csv files.

tooluniverse-antibody-engineering