meta-dream-nap

Consolidates session learnings into the current project's memory chain after context compaction.

1|Updated Jun 23, 2026
One-click install
npx skills add https://github.com/bitranox/bitranox-skills --skill meta-dream-nap-bitranox
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: meta-dream-nap
Source: https://github.com/bitranox/bitranox-skills/tree/main/plugins/bitranox/skills/meta-dream-nap
Command: npx skills add https://github.com/bitranox/bitranox-skills --skill meta-dream-nap-bitranox

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a Claude Code session is compacted or accumulates lessons worth keeping, running the full tree-wide memory consolidation is too expensive, but skipping consolidation loses detail. This Skill performs a fast, budget-bounded consolidation of only the current directory's altitude chain (project, ancestors, anchor) and explicitly reports what it deferred to the full dream. ## Core Features & Use Cases - Capture-first consolidation: Folds session learning signals, audit salvage candidates, and lessons recorded under ## Lessons for the next nap in the repo's handover.md into the persistent memory store. - Chain-scoped passes: Runs backup, toolbox inventory read, dedup, placement, and pruning strictly within the current project's altitude chain, leaving sibling projects and other trees untouched. - Explicit deferred list: Reports exactly which work (sibling dedup, tree-wide placement, descriptor synthesis, toolbox consolidation) is left for the full meta-dream-tree pass. - Use Case: After a context compaction mid-session, invoke a nap to fold the PreCompact hook's salvaged learnings into memory in minutes instead of running the full tree-wide dream. ## Quick Start Ask Claude to run a dream nap to consolidate this session's learnings into the current project's memory chain.

Frequently Asked Questions about meta-dream-nap

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I consolidate session learnings without a full memory dream?▼

Invoke the nap via "nap", "dream nap", or "/dream-nap" to run a quick consolidation of only the current directory's altitude chain. It captures session signals, dedups, places, and prunes within the chain in minutes, then lists what it deferred.

When should I run a nap instead of the full dream tree consolidation?▼

Run a nap around context compactions or when a session accumulated several learning signals but a full dream is not worth its cost. The SessionStart consolidation-due nudge still requires the full meta-dream-tree pass; a nap is partial by design.

Does the nap touch sibling projects or other memory trees?▼

No. Steps 2b through 6 are strictly chain-only and never read or write siblings or other trees. The single exception is capture, which may upsert a fact at the sibling level that owns its pointer, and that write must be named in the report.

What happens to work the nap cannot finish?▼

The nap ends its report with an explicit deferred list for the next full dream, covering sibling dedup, tree-wide placement, descriptor synthesis, behavioral passes, queued skill or hook contributions, and the toolbox consolidation pass.

Why should I not run dream_state.py done after every nap?▼

Running dream_state.py done is only correct when the triggering nudge asked for a nap. If a consolidation-due nudge fired, marking done silences it without the tree-wide work happening, leaving the store partially consolidated.