The Agent Skills Directory

[DYNAMIC_EXECUTION]: The references/workflows.md file includes a Python code example for caching analysis results that uses pickle.load(). The pickle module is insecure for deserializing data from files as it can lead to arbitrary code execution if the file content is untrusted.
[INDIRECT_PROMPT_INJECTION]: Several scripts within the skill ingest external biological data files, creating a potential surface for indirect prompt injection if malicious instructions are embedded in scientific data.
Ingestion points: FASTA files are processed by read_fasta in scripts/batch_sequence_analysis.py, and gene lists are loaded by read_gene_list in scripts/enrichment_pipeline.py.
Boundary markers: The scripts do not use specific delimiters or instructions to ignore potential commands embedded within the data files.
Capability inventory: The skill is capable of network communication via API calls, local file system writes, and execution of bioinformatics tools through the gget library.
Sanitization: The data parsing logic lacks validation or sanitization of the input content before it is processed by the agent or passed to tools like gget gpt.
[EXTERNAL_DOWNLOADS]: As described in SKILL.md and the reference documentation, the gget setup command facilitates the download of large external datasets and models, such as AlphaFold parameters and the ELM database, from official bioinformatics repositories to the local environment.

gget