The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the processing of untrusted biological sequence files. \n- Ingestion points: External files (FASTA, FAA, FNA, FFN, GenBank) are ingested by scripts/curate_fasta.py and analyzed using shell tools like grep and awk as described in SKILL.md. \n- Boundary markers: There are no explicit boundary markers or instructions to the agent to disregard potential instructions embedded within sequence headers or comments. \n- Capability inventory: The skill allows the agent to execute shell commands (via uv run, grep, awk) and perform file system writes (curated database, mapping tables, reports). \n- Sanitization: While scripts/curate_fasta.py contains a normalize_header function that cleans headers for the final output, the agent is instructed to inspect raw headers (e.g., grep "^>" database.fasta | head -20) which could contain injection payloads.

bio-fasta-database-curator