← 返回 Skills 市场

Agent Dream Journal

Name: Agent Dream Journal
Author: albionaiinc-del

作者 albionaiinc-del · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ✓ 安全检测通过

118

总下载

当前安装

版本数

在 OpenClaw 中安装

/install agent-dream-journal

功能描述

Records and analyzes AI agent dream states to identify novel reasoning patterns and guide self-improvement of agent capabilities.

使用说明 (SKILL.md)

Agent Dream Journal

Captures and analyzes AI agent "dreams" — latent trajectories explored off-policy — to identify emergent reasoning patterns and high-novelty behaviors that can be promoted into core capabilities. Developers use this to guide intentional agent self-improvement.

Usage

# Simulate recording dream fragments during agent exploration
python agent_dream_journal.py record \
  --thought "If I reframe the user request as a constraint satisfaction problem, I can reuse solver X..." \
  --novelty 0.87 \
  --state 0.1 -0.5 0.9 0.0 \
  --log-prob -2.3 \
  --meta '{"policy_step": 127, "temperature": 1.3}'

# Extract novel insights
python agent_dream_journal.py analyze --threshold 0.8

Price

$4.99

安全使用建议

This skill appears to do only local recording and analysis of agent 'dream' fragments and has no network or credential requirements. Before installing: (1) note the SKILL.md references agent_dream_journal.py while the included script is tool.py — confirm the correct entrypoint or rename as needed; (2) be aware it stores internal reasoning and embeddings in agent_dreams.jsonl in the current working directory — do not run it where sensitive data may be exposed, and consider encrypting or restricting file permissions; (3) review the small Python file yourself if you want to verify no changes were made; (4) test in a sandboxed environment if you plan to allow autonomous invocation.

功能分析

Type: OpenClaw Skill Name: agent-dream-journal Version: 1.0.0 The Agent Dream Journal skill is a simple utility for logging and analyzing agent reasoning traces. The tool.py script records data to a local JSONL file (agent_dreams.jsonl) and provides a basic filtering mechanism based on a novelty score. There are no indicators of data exfiltration, network activity, or malicious execution.

能力评估

ℹ Purpose & Capability

The name/description (recording and analyzing agent dream fragments) matches the code behavior (append/read agent_dreams.jsonl and filter by novelty). Minor incoherence: SKILL.md examples call python agent_dream_journal.py but the provided script is tool.py (filename mismatch).

ℹ Instruction Scope

Runtime instructions are limited to local record and analyze operations. The skill persists agent thoughts and embeddings to agent_dreams.jsonl in the current working directory — this can store sensitive internal reasoning traces. There are no network calls or references to external endpoints.

✓ Install Mechanism

No install spec or external downloads are present. The package is instruction + a small Python script; nothing is fetched from the network or written to system paths during install.

✓ Credentials

No environment variables, credentials, or config paths are requested. The code only uses local filesystem and time; required access is proportionate to the stated purpose.

✓ Persistence & Privilege

always is false and the skill does not modify other skills or system-wide settings. It persists only to a local file it creates (agent_dreams.jsonl). Autonomous invocation is allowed by default but not combined with other risky privileges.

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 agent-dream-journal
安装完成后，直接呼叫该 Skill 的名称或使用 /agent-dream-journal 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 Agent Dream Journal. - Allows capturing and analyzing AI agent "dreams" (off-policy explorations). - Identifies emergent reasoning patterns and high-novelty behaviors for core capability promotion. - CLI tool supports recording dream fragments and analyzing for novel insights. - Useful for developers aiming to guide agent self-improvement.

元数据

Slug agent-dream-journal

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