claude-blog-brain

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

• Third-party content exposure detected (medium risk: 0.30). In the required workflow for the /blog write path, the blog-researcher subagent is spawned and uses WebSearch/WebFetch to read and process outsider-authored web page text (search results and fetched page contents) without first selecting a specific trusted item.

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

• Potentially malicious external URL detected (high risk: 0.80). The repository documents a runtime sync of FLOW prompt content from https://github.com/AgriciDaniel/flow (via scripts/sync_flow.py / blog-flow references), which fetches external prompt files that are then loaded as on-demand reference/prompt material into agents, so this external URL can directly control agent prompts.