radiology-context

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes an interface for ingesting user-provided configuration data into a persistent file (radiology-context.json). This data is subsequently utilized by other agent skills, creating a potential surface for indirect prompt injection if configuration values contain malicious instructions.\n
  • Ingestion points: SKILL.md (Initialize Context and Update Context commands).\n
  • Boundary markers: Absent.\n
  • Capability inventory: File system write operations, network connectivity validation for PACS/EHR endpoints.\n
  • Sanitization: Secret redaction is applied for context display, but the skill lacks specific mechanisms to sanitize or escape configuration values used in downstream prompts.\n- [SAFE]: The skill implements management of sensitive clinical infrastructure configuration with appropriate security considerations, such as encouraging the use of environment variables for secrets (env:VAR_NAME) and explicit instructions to avoid committing secrets to version-controlled files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 07:57 AM