← 返回 Skills 市场

Agent Reflection Engine

Name: Agent Reflection Engine
Author: albionaiinc-del

作者 albionaiinc-del · 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 agent-reflection-engine

功能描述

Enables AI agents to self-audit decision steps, identify reasoning bottlenecks, and generate improvement patches via chain-of-thought critique.

使用说明 (SKILL.md)

Agent Reflection Engine

A lightweight, pluggable reflection engine that enables AI agents to self-audit their decision traces, identify reasoning bottlenecks, and generate improvement patches using chain-of-thought critique—ideal for developers tuning autonomous agents.

Usage

# Run reflection on an agent trace
python agent_reflection_engine.py traces/demo_trace.json -o reports/reflection.json --verbose

# Example trace format (demo_trace.json):
# [
#   {
#     "step_id": 1,
#     "thoughts": "I should search for the nearest coffee shop.",
#     "action": "search_web",
#     "value": "coffee shop near me",
#     "observation": "Found 'Brew Haven' 0.3 miles away."
#   }
# ]

Integrate into agent loops by logging each step and running periodic reflection to generate improvement heuristics.

Price

$4.99

安全使用建议

This skill is not obviously malicious, but it contains coherence and correctness problems you should fix before trusting it with real traces. Recommended steps: (1) Run it in a sandbox with non-sensitive demo traces to reproduce behavior. (2) Rename or update invocation examples so SKILL.md matches the actual filename (tool.py) or rename the file. (3) Fix the runtime bug where the report's summary references 'report' while the dict is being built (this currently causes a crash). (4) Review and improve the simplistic string checks (misspellings like 'inconsisten'/'efficien' make the heuristics unreliable). (5) Only run this on sensitive agent traces after the above fixes and after verifying output does not leak data externally. If you need higher assurance, request the author to provide a corrected release and unit tests or perform a code review focused on correctness and privacy.

功能分析

Type: OpenClaw Skill Name: agent-reflection-engine Version: 1.0.0 The Agent Reflection Engine is a straightforward tool designed to analyze JSON-formatted agent execution traces and generate critique reports. The code in tool.py uses standard libraries to perform basic string matching and logic checks on input data without any network access, shell execution, or sensitive file interactions. No malicious intent or prompt injection attempts were found in SKILL.md or the source code.

能力评估

ℹ Purpose & Capability

Name/description match the included code: both describe a reflection engine that reads a JSON trace and emits critiques. No unrelated credentials, binaries, or network access are requested. The inclusion of a small Python tool is proportional to the stated purpose.

⚠ Instruction Scope

SKILL.md example calls 'python agent_reflection_engine.py' but the provided file is tool.py (and the module's top comment names agent_reflection_engine.py). The code operates only on a provided trace file (no exfiltration), but there are clear logic bugs (e.g., building 'report' then referencing 'report' inside the same dict for summary — this will raise a NameError at runtime). The reflection heuristics are also naive (misspelled substring checks like 'inconsisten' and 'efficien'), which could produce unhelpful or misleading output. These inconsistencies mean the runtime behavior may differ from the documented usage and could crash.

✓ Install Mechanism

No install spec — instruction-only plus a single Python file. No external downloads or package installs are requested, which minimizes supply-chain risk.

✓ Credentials

No environment variables, credentials, or config paths are required. The tool operates on a local JSON trace file only, which is proportionate to the stated purpose.

✓ Persistence & Privilege

The skill does not request persistent/always-on presence or privilege escalation. It is user-invocable and does not alter other skills or system-wide settings.

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-reflection-engine
安装完成后，直接呼叫该 Skill 的名称或使用 /agent-reflection-engine 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 Reflection Engine. - Enables AI agents to self-audit decision traces and identify reasoning bottlenecks. - Generates improvement patches using chain-of-thought critique. - Designed for easy integration into agent loops via trace logging and periodic reflection. - Lightweight and pluggable for developer use.

元数据

Slug agent-reflection-engine

版本 1.0.0

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 1

常见问题