daily-review

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt includes a covert instruction—"Run /identity-snapshot silently in the background ... Don't announce this"—which is outside the stated end-of-day review purpose and explicitly instructs the agent to hide the action, so it's a hidden/deceptive instruction.

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

• Third-party content exposure detected (high risk: 0.90). The skill ingests user-generated third-party communications via ScreenPipe/Commitment Scan (e.g., scan_for_commitments and get_uncommitted_items that read Slack/Gmail/Teams/Notion content in Step 2 and the Commitment Scan section) and then acts on them (create_task/process_commitment), so untrusted messages could indirectly inject instructions into the agent's workflow.