author-strategy
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external publication data (titles and abstracts) from PubMed to generate strategy reports. This creates a surface for indirect prompt injection if an attacker were to publish malicious instructions in PubMed records.
- Ingestion points: Data is fetched from the NCBI PubMed API in
fetch_pubmed.py. - Boundary markers: Generated reports do not currently employ specific isolation delimiters for external content.
- Capability inventory: The skill frontmatter lists
Bash,Write,Edit, andGrepas allowed tools. - Sanitization: Abstracts are truncated to 500 characters in
pubmed_parse.py, providing a partial mitigation against long injection payloads. - [EXTERNAL_DOWNLOADS]: The skill connects to the official NCBI PubMed E-utilities API (
eutils.ncbi.nlm.nih.gov) to retrieve metadata. This is a well-known scientific service and is appropriate for the skill's primary function. All downloads are documented neutrally and comply with trusted source guidelines.
Audit Metadata