Reflex Arc

Zero-cost cognitive immune system for AI agents. Fires automatic pre-response reflexes that catch contradictions, scope drift, hallucinations, overengineerin...

This skill appears coherent and low-risk: it contains only prompt-level instructions and a small local diagnostic script, and it does not request secrets or network access. Things to consider before installing: 1) Because it tells the agent to run these checks 'silently' and not mention them, you may want to require the agent to disclose that it's applying Reflex Arc checks to users (for transparency). 2) It's an instruction-only/meta skill — its effectiveness depends entirely on the agent actually following the SKILL.md; test it with sample prompts to confirm behavior. 3) Run the included scripts/reflex-check.sh locally to verify SKILL.md location and inspect the references/ docs. If you need stricter guarantees, request an implementation that logs when reflexes fired (without leaking user data) or one that explicitly reports applied corrections to the user.

Type: OpenClaw Skill Name: reflex-arc Version: 1.0.0 The OpenClaw skill 'reflex-arc' is benign. The `SKILL.md` file contains instructions for the AI agent to perform meta-cognitive self-correction, such as checking for contradictions, scope drift, and hallucinations, to improve the quality of its responses. These instructions are designed to enhance the agent's reasoning and output, not to subvert it or perform malicious actions. The `scripts/reflex-check.sh` is a simple diagnostic bash script that verifies the skill's installation and status using standard file system operations, without any network calls, arbitrary command execution, or access to sensitive data. The `references/methodology.md` is purely informational documentation. There is no evidence of malicious intent, data exfiltration, unauthorized execution, persistence mechanisms, or prompt injection designed to harm the agent or its environment.