The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/workflow_generator.py script utilizes f-string templates to construct bash workflow scripts based on user-supplied parameters like file paths and thread counts. This dynamic assembly of executable shell scripts is a common pattern in scientific computing but represents a runtime code generation surface.
[COMMAND_EXECUTION]: Documentation in SKILL.md instructs the user or agent to modify file permissions using chmod +x on generated scripts and subsequently execute them (e.g., ./qc_workflow.sh).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process genomic data files from potentially untrusted external sources (BAM, BED, bigWig).
Ingestion points: The scripts/validate_files.py script reads and parses BED files to validate coordinate integrity, and various tools described in references/tools_reference.md process binary and tabular genomic data.
Boundary markers: There are no explicit instructions or delimiters used when the agent is prompted to process external data files to ignore embedded natural language instructions.
Capability inventory: The skill facilitates extensive file system access and shell command execution via the generated bioinformatics workflows.
Sanitization: While scripts/validate_files.py performs basic format validation (e.g., verifying integer coordinates), it does not sanitize metadata or comment fields in genomic files for potential malicious instructions targeting an LLM context.

deeptools