deep-investigate

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill requires subagents to include "concrete evidence" and verbatim code excerpts from repository files (path:line, short code excerpts) without any instruction to redact or avoid secrets, so any API keys/passwords present in those files would be output verbatim — an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.75). The skill’s Phase 1 “external topics” path explicitly performs runtime WebSearch/WebFetch, and the resulting fetched page text can be ingested into the LLM context to ground the outline and then passed as anchors to Phase 3 subagents (outsider-authored public web content).