openclaw-superpowers
A library of 56 skills for the OpenClaw self-modifying AI agent runtime, adding structured planning, security guardrails, cron-scheduled maintenance, crash recovery and the ability to write its own new skills.
openclaw-superpowers is a skill library for OpenClaw, an open-source AI agent runtime that runs persistently rather than only responding inside a chat window. Installing its 56 skills in one command gives an OpenClaw agent brainstorming, planning and systematic-debugging skills before it acts; six security skills that detect prompt injection, block dangerous actions, audit installed code and scan for leaked credentials; twelve cron-scheduled skills for memory cleanup, health checks, budget tracking and community monitoring; and self-recovery, loop-breaking and task-handoff skills for crashes and restarts.
Its most distinctive skill lets the agent write a new skill file during an ordinary conversation, so a stated preference such as always checking for security issues in a review becomes a permanent, live behavior with no restart needed. A separate skill scans Reddit periodically for community pain points and feature requests and scores them by signal strength.
It is inspired by the unrelated obra/superpowers project, extended for agents meant to run continuously.