support-feedback-prioritization

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly asks to include the "raw text" of Intercom and Slack signals in the report (and does not instruct redaction), so any credentials or tokens present in user messages would be reproduced verbatim by the LLM even though service auth itself is handled externally.

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly pulls user-generated content from Intercom tickets and Slack channels (see "Step 1 — Collect signals" and the MCPs required list) and requires the agent to read raw text and use it to classify, enrich, score, and decide actions, so untrusted third-party content could materially influence behavior.