fix-sentry-issues

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to pull event-level data (Phase 3a) and in Phase 3e to "fetch() the actual URLs that timed out" and use Sentry event fields (URLs, request parameters, stack traces) — i.e., ingest and act on arbitrary external/user-generated URLs and content from Sentry events — which clearly exposes the agent to untrusted third-party content.