funding-trend-forecaster
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes research abstracts, which are considered untrusted external data.
- Ingestion points: The
TextProcessorandTopicModelerclasses inscripts/main.pyread and tokenize abstract text. - Boundary markers: No specific delimiters are used to wrap ingested text or instruct the processing logic to ignore embedded commands.
- Capability inventory: The skill possesses file writing capabilities in
scripts/main.py(viaexport_report) but does not contain capabilities for network transmission or command execution based on the processed text. - Sanitization: Basic text cleaning and normalization are performed by the
clean_textmethod, which filters special characters and numbers. - [SAFE]: All identified dependencies and libraries are standard tools for data science and NLP. The referenced data sources (NIH, NSF, Horizon Europe) are official and trusted institutional repositories. The vendor resource (aipoch/medical-research-skills) aligns with the stated authorship and infrastructure.
Audit Metadata