The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from an external API and binary HDF5 files, which could potentially contain malicious instructions intended for the agent.
Ingestion points: Remote JSON responses from maayanlab.cloud (Query 1-4) and bulk genomic data from HDF5 files (Query 6).
Boundary markers: Absent; the skill does not use specific delimiters or instructions to prevent the agent from interpreting ingested data as commands.
Capability inventory: The skill performs network operations via requests.get and writes local files using pandas.to_csv across various queries in SKILL.md.
Sanitization: Only basic JSON parsing and HTTP status code validation are performed; no sanitization exists for the content of the expression data or metadata.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download large genomic data matrices (HDF5 format) from an external server.
Evidence: Fetches data from https://maayanlab.cloud/archs4/api/v1 and provides URLs for bulk files like https://maayanlab.cloud/archs4/files/human_gene_v2.6.h5. These sources belong to the Ma'ayan Lab (Mount Sinai), a recognized scientific research organization.

archs4-database