cg1262

Dreaming Skill — Development skill for Claude Code

Development community

Personal Claude Code / Codex CLI skill: non-destructive memory consolidation ("dreaming") over CLAUDE.md/AGENTS.md + recent session transcripts.

How to install Dreaming Skill

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cg1262/dreaming-skill and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Dreaming Skill does

Personal Claude Code / Codex CLI skill: non-destructive memory consolidation ("dreaming") over CLAUDE.md/AGENTS.md + recent session transcripts.

Alternatives in Development

Ccusage — CLI for analyzing Claude Code/Codex usage from local JSONL files 11.8k ★
Careful — Activate destructive command warnings for the session 2.8k ★
Coding Agent Session Search — Unified TUI and CLI to index and search your local coding agent session history across 11+ providers (Codex, C 708 ★

README

dreaming-skill

A personal, cross-tool "dreaming" memory-consolidation skill for **Claude Code** and **Codex CLI**.

What this is

Anthropic's Managed Agents product has a research-preview feature called "dreams": periodically, a background pass reads an agent's accumulated memory plus recent session history and produces a consolidated update — merging duplicate notes, dropping stale/contradicted entries, and folding in new insights.

This repo replicates that *pattern* locally, using tools already on this machine, for the two memory files these CLIs actually maintain:

Tool	Memory file	Skill location (after install)
Claude Code	`CLAUDE.md`	`~/.claude/skills/dreaming/`
Codex CLI	`AGENTS.md`	`~/.codex/skills/dreaming/`

Claude Code's current docs say it reads `CLAUDE.md`, not `AGENTS.md`. If a repository already uses `AGENTS.md` for other coding agents, create a `CLAUDE.md` that imports it (`@AGENTS.md`) or symlink `CLAUDE.md` to `AGENTS.md`; the Claude dreaming skill still writes a `CLAUDE.dream.*.md` proposal rather than treating `AGENTS.md` as additional native Claude memory.

How it differs from the real "dreams" feature

Not the same feature, not calling that API. This is a locally-run skill/prompt, invoked manually (or via a slash command / natural-language request), not a managed background job running on Anthropic's infrastructure.
Never touches the original memory file. The real dreams feature (per its research-preview description) also produces a proposal for review, but here that's an absolute, hard-coded rule: the skill is instructed to never write to CLAUDE.md/AGENTS.md, only to a new CLAUDE.dream..md / AGENTS.dream..md file. You review it and then adopt it yourself — with scripts/promote_dream.sh (see Adopting a dream) or by hand — or delete it.