ce-sweep

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to "seed" subagents with the source's config entry verbatim, which forces the LLM to include any config-stored credentials or tokens verbatim in generated prompts/requests to those subagents (an exfiltration risk).

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow dispatches extraction and media-analysis subagents in Phase 2b/2e that ingest outsider-authored free text from configured feedback sources (e.g., Slack messages, GitHub issues, and email bodies/attachments) and process their body/transcript content as DATA.