The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill communicates with the external NCBI BLAST service at blast.ncbi.nlm.nih.gov. While this is the intended primary function of the bioinformatics tool, the domain is not on the predefined trusted list.
[DATA_EXFILTRATION] (LOW): The skill transmits biological sequence data provided via the --sequence or --fasta arguments to the NCBI servers for processing. Users should ensure they are not submitting proprietary or sensitive genomic data if they wish to remain private.
[COMMAND_EXECUTION] (SAFE): The provided Python scripts (run_blast_biopython.py, etc.) utilize well-maintained and reputable libraries such as biopython and typer. The code does not perform any unauthorized system commands or privilege escalation.
[PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection (Category 8). It retrieves data from an external, potentially attacker-influenced source (BLAST hits from NCBI) and returns this data in a structured format. If a downstream LLM processes the hit titles or descriptions without sanitization, it could be influenced by malicious content embedded in those strings.
Ingestion points: Untrusted sequence data enters through the --sequence CLI argument or the contents of files specified by --fasta in scripts/run_blast_biopython.py.
Boundary markers: Absent; sequence data and resulting BLAST hits are not enclosed in delimiters or accompanied by instructions to ignore embedded commands.
Capability inventory: The script performs network requests to NCBI via NCBIWWW.qblast and can write results to the local filesystem using Path.write_text if an --output path is provided.
Sanitization: No sanitization is performed on the title or accession fields retrieved from NCBI before they are serialized into JSON output.

bio-blast