The Agent Skills Directory

[SAFE]: No malicious patterns or safety violations detected. The skill's behavior aligns with its described purpose of identifying metabolites and searching scientific repositories.
[PROMPT_INJECTION]: Potential for indirect prompt injection exists because the skill ingests untrusted data from user queries and external scientific databases.
Ingestion points: User-provided metabolite_list, study_id, and search_query in SKILL.md.
Boundary markers: The skill instructions do not specify any delimiters or safety prompts to isolate external data from the execution logic.
Capability inventory: The skill utilizes tools to query external APIs (HMDB, PubChem, MetaboLights) and writes research results to a local file using the output_file parameter.
Sanitization: No input validation or output sanitization routines are specified to mitigate potential injection within generated markdown reports.

tooluniverse-metabolomics