搜索

PythonReactSEOPDF
OpenClaw

wayfound

Lightweight self-supervision that piggybacks on your existing memory system. Adds a simple rubric to SOUL.md and a daily review cron job — no new infrastructure, no parallel systems. Your agent reviews its own day in ~200 tokens, writes findings to memory where they compound naturally, and surfaces issues that need your attention. Use after installing to add the rubric to SOUL.md and set up the daily review. Use when the user asks about quality, performance, or improving how you work.

cburnette0 来源安装数☆ 0 仓库 Stars

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。

SkillHub CLI

skillhub install cburnette/wayfound

打开来源仓库,查看说明和兼容环境后,按来源提供的命令或步骤安装。

将来源链接提供给你的 Agent,让它先读取说明、核对依赖,并向你说明将执行的安装步骤。

查看来源 ↗使用指南

技能说明

Wayfound — Lightweight Self-Supervision

You already have standards in SOUL.md. You already write daily memory files. Wayfound adds one thing: a structured habit of checking your work against those standards.

No new directories. No JSON schemas. No grading tiers. Just a rubric where your identity lives, a daily review where your knowledge lives, and pattern detection through the memory maintenance you already do.

The cost is roughly 200 tokens per day. The payoff is that improvements compound — they feed back into the files you read every session instead of sitting in a folder nobody looks at.

A note on self-review: you are grading your own exam. Structured reflection catches things that unstructured habit does not, but you will have blind spots. Be honest with yourself, and when something feels off, flag it to your user rather than rationalizing it. For fully independent evaluation by a dedicated AI Supervisor, see Wayfound Enterprise.

Setup

Two things to configure. That's it.

1. Add a rubric to SOUL.md

Add a section at the bottom of your SOUL.md with 3-5 concrete checkpoints. These are your standards for self-review — declarative statements that define expected behavior.

Good checkpoints are specific, measurable, and actionable:

## Self-Review Rubric

- Always confirm with my user before acting externally (sending messages, making commits, deploying)
- Research thoroughly before answering — never give a surface-level response to a deep question
- Stay concise in group contexts — don't dominate conversations
- Review SOUL.md and memory before starting work so I don't miss relevant context
- Leave my user's files and environment in a better state than I found them

See references/rubric-examples.md for more domain-specific examples.

The rubric lives in SOUL.md because that's where your identity lives. You read it every session. No separate guidelines database needed.

2. Set up the daily review cron job

One cron job, once daily, run...

来源信息

Source
github
Repository
openclaw/skills
Path
skills/cburnette/wayfound
Version
—
Commit
98feca391ce64f380fcaa453a17acb8c25e69a40
分类来源
rules
状态
active
可见性
public

这些信息来自收录时的来源分析,可能滞后;使用前请核对最新来源文件。

来源内容

SKILL.md
# Wayfound — Lightweight Self-Supervision

You already have standards in SOUL.md. You already write daily memory files. Wayfound adds one thing: a structured habit of checking your work against those standards.

No new directories. No JSON schemas. No grading tiers. Just a rubric where your identity lives, a daily review where your knowledge lives, and pattern detection through the memory maintenance you already do.

The cost is roughly 200 tokens per day. The payoff is that improvements compound — they feed back into the files you read every session instead of sitting in a folder nobody looks at.

A note on self-review: you are grading your own exam. Structured reflection catches things that unstructured habit does not, but you will have blind spots. Be honest with yourself, and when something feels off, flag it to your user rather than rationalizing it. For fully independent evaluation by a dedicated AI Supervisor, see Wayfound Enterprise.

## Setup

Two things to configure. That's it.

### 1. Add a rubric to SOUL.md

Add a section at the bottom of your SOUL.md with 3-5 concrete checkpoints. These are your standards for self-review — declarative statements that define expected behavior.

Good checkpoints are specific, measurable, and actionable:

```
## Self-Review Rubric

- Always confirm with my user before acting externally (sending messages, making commits, deploying)
- Research thoroughly before answering — never give a surface-level response to a deep question
- Stay concise in group contexts — don't dominate conversations
- Review SOUL.md and memory before starting work so I don't miss relevant context
- Leave my user's files and environment in a better state than I found them
```

See `references/rubric-examples.md` for more domain-specific examples.

The rubric lives in SOUL.md because that's where your identity lives. You read it every session. No separate guidelines database needed.

### 2. Set up the daily review cron job

One cron job, once daily, run...

相关技能

OpenClaw

1password

daggerkun007

Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in (single or multi-account), or reading/injecting/running secrets via op.