monitor

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md and workflow clearly show the system fetching public, user-generated sources (atom/rss feeds, GitHub repos, arbitrary pages, and Bluesky/AT protocol handles) via the feed-checker and subscription/check-feeds pipeline (packages/.../feed-checker.ts and subscription/summarize) and passing detected entries to an LLM summary that drives notifications and publish/dismiss actions, so untrusted third-party content can materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches external feeds at runtime (e.g., https://simonwillison.net/atom/everything/) and injects that fetched content into the LLM summarization step, so remote content can directly influence agent prompts/outputs.