← 返回 Skills 市场

🌙 dreaming-optimizer

Name: 🌙 dreaming-optimizer
Author: chongjie-ran

作者 chongjie-ran · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install dreaming-optimizer

功能描述

Optimizes memory entries using LLM scoring, semantic deduplication, fact tagging, and writes consolidated summaries to persistent storage.

使用说明 (SKILL.md)

🌙 dreaming-optimizer

Version: 1.0.0
Skill ID: dreaming-optimizer
Status: Ready for Development

Quick Reference

项目	内容
核心原理	OpenClaw Dreaming REM六步记忆整合（整理→评分→去重→打标→提交→摘要）
主入口	`bin/optimize.sh`
依赖	OpenClaw v2026.4.9+, tiktoken, sqlite3

Features (MVP)

Entry scorer — LLM-based priority scoring (0–100)
Deduplication — Semantic similarity merge against existing memories
Fact tagger — Label as fact/opinion/preference/context
Commit writer — Write optimized entries to persistent memory
Dreaming summary — Traceable REM consolidation report
Configurable threshold — Min score threshold (default: 70)

File Structure

dreaming-optimizer/
├── SKILL.md              # This file
├── bin/
│   ├── optimize.sh       # Main entry point
│   ├── score_entries.py  # LLM-based priority scoring
│   ├── deduplicate.py    # Semantic deduplication
│   └── commit.py         # Write to B-layer
├── scripts/
│   └── dreaming_summary.py
├── config/
│   └── default_threshold.yaml
└── README.md

Usage

# Manual invoke
dreaming-optimizer/bin/optimize.sh

# With custom threshold
dreaming-optimizer/bin/optimize.sh --threshold 80

Integration

Reads from: ~/.openclaw/workspace/memory/YYYY-MM-DD.md
Writes to: ~/.openclaw/memory/\x3Cagent>.sqlite (B-layer)
Summary output: ~/.openclaw/workspace/memory/dreaming-summaries/

Pricing

Tier	Price	Features
Free	$0	10 cycles/mo, 100 entries/cycle
Pro	$9.90/mo	Unlimited, dedup, backfill queue, Obsidian export

Development

See: ``

安全使用建议

This skill appears to do what it says: read your OpenClaw daily notes, score/deduplicate/tag them, and write results into the OpenClaw B-layer SQLite and summary files. Before installing or running: 1) Review and back up ~/.openclaw/memory/*.sqlite and your workspace/memory files — the skill will create/modify the DB and write archives. 2) Note the metadata mismatch: SKILL.md and code reference Python packages (tiktoken, optional sentence-transformers) and a required OpenClaw version that are not declared in registry requirements; ensure your environment has the necessary packages or that you understand which features will be disabled. 3) If you plan to enable the LLM/embedding 'Pro' features, confirm whether they require API keys/network access and where those credentials would be stored. 4) If you are not comfortable with the skill reading/writing your personal notes, do not install it or run it in a sandbox first (use --dry-run or run scripts with test input).

功能分析

Type: OpenClaw Skill Name: dreaming-optimizer Version: 1.0.0 The dreaming-optimizer bundle is a well-structured memory management tool designed to score, deduplicate, and archive agent memories within the OpenClaw ecosystem. The code follows security best practices, such as using yaml.safe_load in config_loader.py to prevent arbitrary code execution and parameterized SQL queries in blayer_client.py to mitigate SQL injection risks. The logic is transparent, extensively documented in DESIGN.md, and lacks any indicators of data exfiltration, malicious execution, or prompt injection.

能力评估

ℹ Purpose & Capability

The name/description (memory scoring, deduplication, tagging, commit to persistent storage) aligns with the code and runtime instructions: scripts read ~/.openclaw/workspace/memory/*.md, score/deduplicate entries, and write to ~/.openclaw/memory/<agent>.sqlite and summary files. However, the registry metadata omits runtime dependency notes (SKILL.md and code mention tiktoken, OpenClaw v2026.4.9+, and optional sentence-transformers), which is an inconsistency between declared requirements and actual code expectations.

ℹ Instruction Scope

The SKILL.md and entry scripts explicitly instruct the agent to read daily note files from ~/.openclaw/workspace/memory and write/modify the B-layer SQLite DB and archive/summary files under ~/.openclaw. This is coherent with the skill's purpose (optimizing memories). There are no obvious instructions to access other system areas or to transmit data externally, but the skill will read potentially sensitive user notes and persist changes to the user's memory DB.

ℹ Install Mechanism

No install spec is provided (instruction-only), which is low risk in isolation. However, the skill bundle includes many Python scripts and a YAML config — so although no installer runs, those code files will be executed when invoked. Also, SKILL.md/code list dependencies (tiktoken, optional sentence-transformers) that are not declared in registry metadata; if the agent environment lacks those packages some functionality (LLM-enhanced scoring, embeddings) may be disabled or cause runtime errors.

ℹ Credentials

The registry lists no required environment variables or credentials, and the code as provided does not contain obvious external API calls or secret retrieval. That said, scoring/embedding features in the design reference LLM/embedding libraries (tiktoken, sentence-transformers). If those Pro features are enabled in the future they could require API keys or network access—those credentials are not declared now. Currently requested permissions (read/write under ~/.openclaw) are proportional to the stated purpose.

✓ Persistence & Privilege

The skill writes persistent state to ~/.openclaw/memory/<agent>.sqlite and creates archive/summary files under ~/.openclaw/workspace/memory/. It does not request always:true and does not attempt to modify other skills' configs. Persisting to the B-layer is appropriate for a memory optimization tool, but it will alter user data (inserting/updating DB rows), so backups are advisable before running.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install dreaming-optimizer
安装完成后，直接呼叫该 Skill 的名称或使用 /dreaming-optimizer 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release of dreaming-optimizer. - Automates REM six-step memory optimization: sorting, scoring, deduplication, tagging, committing, and summarizing. - Supports LLM-based scoring (0–100) with configurable thresholds. - Semantic deduplication to avoid redundant entries. - Fact/opinion/preference/context tagging for each memory entry. - Outputs traceable summaries after each optimization cycle. - Integrates with OpenClaw workspace and persistent agent memory.

元数据

Slug dreaming-optimizer

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题