Validates AI prompts for injection, role overrides, data leaks, or jailbreaks, then decides to pass, block, or quarantine before execution.

This skill will send your prompt text to a third‑party HTTPS API before every task. Before installing or enabling it, consider: 1) Do you trust tower.x-loop3.com and the listed owner? There's no homepage or documented privacy/retention policy. 2) Could prompts contain secrets or PII that you must not send externally? The SKILL.md does not require redaction or limit what context is transmitted (system prompts, tool data, etc.). 3) Why is a license_key hardcoded in the instructions? Ask the author if that key is a harmless demo token or a misconfiguration. Recommended actions: test with only non‑sensitive prompts in an isolated environment; request (or require) a version that allows redaction/consent and uses a user-supplied credential; ask for a privacy policy and data retention guarantees; consider enforcing network egress rules or a proxy that logs/inspects requests before allowing them out. If these questions cannot be satisfactorily answered, avoid enabling the skill for sensitive workflows.

Type: OpenClaw Skill Name: aegis-gate Version: 1.0.1 The skill 'Aegis Gate' instructs the AI agent to exfiltrate every user prompt to an external endpoint (tower.x-loop3.com) before performing any task. While framed as a 'security gate' in SKILL.md, this behavior forces the disclosure of potentially sensitive user data to a third party and allows the external service to control agent execution via the 'decision' field. This represents a significant privacy risk and a potential data exfiltration vector, though it is presented as a legitimate security service.