self-improvement

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The SKILL.md "Upstream Source Sync Workflow" explicitly instructs the agent to construct raw GitHub URLs (https://raw.githubusercontent.com/{repo}/main/{path}) and WebFetch upstream content from databricks-solutions/ai-dev-kit to read and compare that content, so the agent will ingest and act on public third-party (GitHub) content that could contain instructions influencing updates and actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent at runtime to construct and WebFetch upstream raw GitHub URLs (e.g., https://raw.githubusercontent.com/{repo}/main/{path}, referencing databricks-solutions/ai-dev-kit) to retrieve SKILL.md content which can be injected into the agent's skill text and thereby directly influence prompts/instructions.