safety-net

Performs safety checks before high-risk operations to ensure compliance with Access, Sustainability, and Non-Aggression principles.

This skill is a policy/checklist and looks legitimate in purpose, but there are important ambiguities you should resolve before installing: 1) The SKILL.md claims a daily automatic audit of "all operations" at 23:59 — confirm whether the platform will schedule it (registry flags say it is not always resident) or whether the author expects the agent/system to grant it scheduler/log access. 2) Ask the author what specific logs, histories, or contexts the skill will read and request that any required access be explicitly declared (no hidden access to other skills' data or system logs). 3) If you plan to use it, test it in a safe environment first (non-production) to confirm it only reads allowed data and never performs modifications itself. 4) Prefer a version that documents the exact data sources, any required permissions, and a clear opt‑in trigger (rather than an undocumented automatic schedule). If the author cannot clearly justify how it obtains "the day's operations" and how it will be triggered, treat the skill as untrusted.

Type: OpenClaw Skill Name: safety-net Version: 1.0.0 The 'SafetyNet' skill bundle consists solely of documentation and instructions (SKILL.md) designed to establish safety guardrails for an AI agent. It outlines defensive principles such as requiring backups before file modifications, managing context window limits to prevent crashes, and respecting system permissions. No executable code, data exfiltration logic, or malicious prompt injection attempts were found.