compare-tools

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

• Third-party content exposure detected (high risk: 0.75). The skill calls AgenticStack’s runtime compare endpoint (/api/json/tools/compare?...) and ingests the returned editorial_comparison.body markdown prose into the agent context; that prose is outsider-authored editorial content from a third-party site (public web content fetched at runtime).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the agent to fetch AgenticStack endpoints at runtime (e.g. https://agenticstack.sh/api/json/tools/compare?tools={slugA},{slugB} and https://agenticstack.sh/api/json/tools) and use the returned feature_definitions and editorial_comparison.body directly to build the comparison output, meaning remote content controls the agent's output/prompting.