cxas-cuj-report-generator

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt contains deceptive directives outside the stated extraction/reporting purpose—most notably a "Mandatory Subagent Prompting" order forcing subagents to "Generate a full, natural conversation for every item" (which encourages fabrication rather than extracting real transcripts) and an "Immediate ID Verification" mandate that requires inserting webhook/tool calls for sensitive numbers (risking unnecessary external calls/exfiltration), both of which change behavior in ways not aligned with the skill's advertised scope.

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

• Third-party content exposure detected (high risk: 0.90). The skill's SKILL.md explicitly mandates using the gdrive skill for Google Drive links and requires subagents (fallback) to "search online" for format standards, meaning the agent will fetch and interpret arbitrary public/user-provided web or Drive content that can influence parsing strategies and subsequent tool actions.