The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documentation in references/workflow_guide.md and references/pipeline_steps.md suggests using pickle.load() to restore analysis states (DeseqDataSet) or load datasets from disk.
Evidence: with open("counts.pkl", "rb") as f: counts_df = pickle.load(f) and with open("dds_result.pkl", "rb") as f: adata = pickle.load(f).
Risk: Python's pickle module is not secure. Deserializing data from an untrusted source can allow an attacker to execute arbitrary code on the system.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in CSV, TSV, and AnnData formats.
Ingestion points: pd.read_csv() calls in scripts/run_deseq2_analysis.py and documentation files.
Boundary markers: None identified; the skill assumes standard biological data formats.
Capability inventory: The skill has Bash access, Read/Write file system permissions, and executes Python scripts that can generate plots and save data.
Sanitization: The script performs data validation (checking for non-negative integers, matching indices), but these are biological validity checks rather than security sanitization. If the output of the analysis is subsequently processed by the agent to make decisions, a malicious file could potentially influence the agent's behavior via the analysis results.

alterlab-pydeseq2