Auto Dream Light

Lightweight, memory-safe Auto Dream workflow for OpenClaw that consolidates recent notes into existing memory files without replacing the user’s current memo...

Security Scan

The skill's instructions, required resources, and behavior are coherent with its stated purpose (conservative, incremental consolidation of local memory files); it requests no credentials, dependencies, or installs.

Purpose & Capability

Name/description claim a conservative memory-consolidation workflow and every declared requirement matches that: no binaries, no env vars, no installs. The files referenced (MEMORY.md, memory/**/*.md, memory/system/auto-dream-log.md) are directly related to the described purpose.

Instruction Scope

Runtime instructions are limited to scanning recent daily logs, extracting items, routing them into existing memory files, appending a dream-log entry, and optionally committing changes. This is within scope, but the spec is deliberately high-level (relies on the agent's judgment for 'candidate' selection and deduplication). The commit step implies use of VCS if present — the skill does not mention push behavior, so users should confirm commit/push settings.

Install Mechanism

Instruction-only skill with no install spec and no code files — nothing will be written or downloaded during installation. This is the lowest-risk install profile.

Credentials

No environment variables, credentials, or config paths are required. The skill only references project-local files under memory/ and MEMORY.md, which aligns with its function.

Persistence & Privilege

always:false and default model invocation are unchanged. The skill does not request persistent elevated privileges or modify other skills. Its intended writes are local memory files and a dream-log, which are appropriate for its goal.

Assessment

This skill appears coherent and low-risk, but before enabling it: (1) run it manually on a copy of your memory directory to verify behavior; (2) review and approve any file changes before committing; (3) check your git configuration so commits won't be automatically pushed to a remote you don't intend; (4) ensure that sensitive secrets are not stored in the memory files the skill will scan; and (5) prefer manual or semi-auto triggers until you trust its extraction/deduplication choices.

A lightweight, memory-safe Auto Dream skill for OpenClaw that works with the memory system the user already has.

What this skill is for

Use this skill when the user wants:

• on-demand memory consolidation
• a lightweight “dream” workflow
• a dream log with execution history
• a gradual path from manual to semi-automatic memory cleanup
• memory consolidation without replacing the current MEMORY.md structure

This skill is intentionally conservative:

• it does not rebuild the memory architecture
• it does not introduce dashboards, indexes, or archives by default
• it does not assume the user's memory follows a template
• it prefers small, explicit, high-value updates

Why install it

• works with existing MEMORY.md and memory/
• makes conservative, incremental updates
• avoids noisy or fake “productivity” output
• keeps runs easy to review and audit
• supports a clean path from manual runs to semi-automatic operation

Core idea

Instead of forcing a new memory system, this skill works with the one the user already has.

Default approach:

1. scan recent daily logs
2. extract durable, high-value information
3. route items to the correct destination
4. update memory incrementally
5. record the run in a dream log
6. return a concise summary to the user

Read these references when needed

• Read references/adapted-plan.md before designing or changing the workflow.
• Read references/manual-run.md when running the workflow manually.
• Read references/semi-auto.md when the user wants a fixed trigger-based flow or wants to prepare for cron later.

Default file roles

Typical targets:

• MEMORY.md → long-term facts, stable preferences, system conclusions, reusable lessons
• memory/projects/** → project-specific context
• memory/system/auto-dream-log.md → dream execution history only
• memory/YYYY-MM-DD.md → raw daily notes; only add a consolidation marker when appropriate

Operating rules

• Preserve the existing memory structure.
• Prefer project files over stuffing everything into MEMORY.md.
• Skip low-value chat, tests, and one-off noise.
• Do not invent memory just to make a run look productive.
• If there is nothing worth consolidating, explicitly skip and say so.

Trigger guidance

Typical trigger phrases:

• "整理记忆"
• "dream now"
• "跑 dream"
• "做一次记忆整理"
• "跑一次适配版 auto dream"

If the user says dream details, show recent dream-log history instead of running a new consolidation.