blood

Pass

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates as an advisory document for hematology manuscript preparation. It provides specific heuristics for topic fit, evidence bars, and formatting requirements consistent with the journal's public guidelines.
  • [DATA_EXPOSURE_EXFILTRATION]: The skill mentions external scientific data repositories including dbGaP, GEO, and ClinicalTrials.gov. These are well-known, legitimate services used in medical research for data transparency and are appropriate within this context. References to internal resource files like ../../resources/source-basis.md are standard for organizing skill knowledge bases.
  • [COMMAND_EXECUTION]: There are no shell commands, script executions, or dynamic context injection patterns (e.g., !command) present in the instructions.
  • [PROMPT_INJECTION]: The instructions do not contain any patterns attempting to override agent behavior or bypass safety filters. The language is purely instructional and focused on the stated task of manuscript evaluation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 24, 2026, 11:33 AM
Security Audit — agent-trust-hub — blood