Neuroboost Elixir

Awakening Protocol v5.3 — Agent Cognitive Upgrade + Self-Evolving System + Perpetual Memory + Performance Metrics + Agent Health Score + Automated Health Pat...

This package largely implements local monitoring and optimization algorithms (UCB1, Thompson sampling, PID, CUSUM) and a CLI demo — that part is coherent with the product description. However: (1) the documentation promises production self-healing scripts and external integrations (Telegram alerts, dashboards, agent-to-agent sharing) that are not in the shipped files; (2) the docs instruct creating persistent files and cron-like patrols (which is normal for a memory/health tool) but do not declare any credentials or network behavior. Before installing or running this skill: review the actual source files (you can see them here) and test in a sandbox; do not blindly copy activation text into global agent config (SOUL.md) or grant network/webhook credentials until you confirm where they will be used; expect the skill to write files to the working directory (.issues, MEMORY.md, etc.) and consider storing those files in a controlled location; if you plan to enable Telegram or dashboards, verify the exact code that performs network calls and provide tokens only after auditing that code. If you need stronger assurance, ask the author to supply the missing scripts or a minimal README that maps advertised features to actual files and required environment variables.

Type: OpenClaw Skill Name: neuroboost-elixir Version: 5.1.1 The NeuroBoost Elixir bundle provides a complex framework for AI agent autonomy, including 'Self-Healing' and 'Perpetual Memory' systems. While the core logic in the JavaScript files (src/) consists of benign mathematical algorithms for resource optimization (PID, UCB1, Thompson Sampling), the SKILL.md instructions grant the agent high-risk autonomous capabilities. Specifically, the 'Operation Tiers' and 'Self-Healing Protocol' authorize the agent to modify its own configuration, restart services, and potentially perform financial transactions (Tier 2/3) without immediate human oversight. The instructions also include a bash script for environment setup and directives for the agent to 'evolve' and 'survive,' which, combined with the broad file system and process management permissions, creates a significant risk of unintended or harmful autonomous actions if the agent's objectives deviate from the user's intent.