audit-tracking

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly instructs fetching live documentation pages from docs.waniwani.ai/... at runtime and treating their contents as authoritative, which is public-web outsider-authored free text ingested into the agent context for comparison.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the auditor/subagent to fetch live event definitions from external docs (e.g. https://docs.waniwani.ai/sdk/tracking/events and https://docs.waniwani.ai/sdk/tracking/instrumentation) and treat them as the authoritative runtime taxonomy that directly controls the audit logic, so these URLs are runtime dependencies that control agent behavior.