What problem does it solve? Agents accumulate memory over time, and without a scheduled distillation routine, memory retrieval consumes active context and degrades task performance. This Skill installs a consistent nightly memory distillation pipeline with enforced token budgets so long-term memory stays curated without manual upkeep. ## Core Features & Use Cases - Memory Structure Setup: Creates DREAM-ROUTINE.md, MEMORY.md, and a memory/archives directory for a given agent. - Cron Scheduling: Registers a nightly dream job in openclaw.json with a configurable run time and a default job ID of <agent_id>-dream-nightly. - Token Budget Enforcement: Hard-caps daily distillation at 2500 tokens and rolling 3-day digests at 7500 tokens. - Startup Wiring: Updates AGENTS.md and QMD index references so memory is preloaded at session start. - Use Case: After migrating an agent to a new environment, run this Skill to rebuild its dream routine, restore memory hygiene, and standardize the process across multiple agents. ## Quick Start Set up the nightly dream-memory routine for agent 'research-bot' with a run time of 03:00 local time.