Token导航 LogoToken导航

Agent Skill · 研究检索

pr-proof-of-work(pr 工作量证明)

pr-proof-of-work 用于处理浏览器自动化、网页检查和页面信息提取,适合在 OpenClaw 中需要让 Agent 打开页面、读取网页或验证前端流程时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

pr-proof-of-work 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为操作浏览器;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 32,076。

总安装 2,590周安装 109GitHub Stars 公开资料未说明下载 907核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:pr-proof-of-work(pr 工作量证明)
来源仓库:https://github.com/newtontech/pr-proof-of-work
安装命令:
openclaw skills install pr-proof-of-work
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install pr-proof-of-work

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

使用 Playwright 浏览器截图生成 TDD 驱动的 E2E 测试证明。适合需要可视化证据验证前端功能实现的场景。pr-proof-of-work 属于研究检索类 Skill,可作为该场景下的辅助能力补充。在修复 bug 或新增功能时可作为 PR 附件提交。需配置浏览器环境与截图存储路径,注意临时文件清理。

可能消耗较多资源,建议在低峰时段执行长时间任务。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install pr-proof-of-work;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

pr-proof-of-work 怎么安装?

在终端执行 openclaw skills install pr-proof-of-work。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

pr-proof-of-work 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

pr-proof-of-work 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

pr-proof-of-work 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

pr-proof-of-work 用的人多吗?

公开数据是累计安装 2,590 次,近一周安装 109 次,下载 907 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

pr-proof-of-work 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
pr-proof-of-work

pr 工作量证明 · 本页

2,590—操作浏览器MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
tdd-e2e-pr-workflow
description
>
Workflow
pick one open issue → write E2E test → implement fix → PR with screenshots.
Triggers
tdd e2e", "screenshot pr", "before after screenshot", "visual pr proof",

TDD E2E PR Workflow

One Issue → One PR with visual proof.

Flow: Select one issue → Write test (BEFORE screenshot) → Fix → Test passes (AFTER screenshot) → PR with screenshot comment.

Golden Rule: Study ≥2 existing PASSING tests before writing any new test. Wrong fixture/selector usage is #1 failure cause.


Phase 0: Select One Issue

Choose an open issue to work on, or pick randomly:

# Option A: List and manually select
gh issue list --repo <owner/repo> --state open --json number,title,body

# Option B: Pick one randomly
gh issue list --repo <owner/repo> --state open --json number,title | \
  jq -r '.[] | "\(.number): \(.title)"' | shuf -n 1
  1. Note the issue number and create a short kebab-case slug from title (e.g. "fix-discard-button")
  2. Create isolated worktree:
   git worktree add .worktrees/fix/<slug> -b fix/<slug>
   cd .worktrees/fix/<slug>
  1. Copy screenshot-reporter.ts into the worktree's e2e/ directory (not tracked on fix branches)
  2. Set screenshot output:
   export E2E_SCREENSHOT_DIR="$(pwd)/e2e-screenshots"

Phase 1: Write Real Browser Test

Step A: Study existing tests (MANDATORY)

Read ≥2 passing E2E tests to understand:

  • Fixture API (withProject, withSession, gotoSession, sdk)
  • Action helpers (waitSessionIdle, runTerminal, openSettings)
  • Selector patterns (data-component, data-slot, ARIA roles)
  • Data seeding (apply_patch, terminal commands, SDK methods)

Step B: Write test with BEFORE screenshot

import { test, expect } from "../fixtures"
import { createScreenshotReporter } from "../screenshot-reporter"

test("feature description", async ({ page, withProject }) => {
  const screenshot = createScreenshotReporter(page, "test-name")

  await withProject(async (project) => {
    // ... setup using project's fixture patterns ...
    await screenshot.captureBefore("initial-state")
    // ... assertions ...
  })
})

Playwright TS transform gotcha — rejects inline object params:

// WRONG: fs.mkdirSync(dir, { recursive: true })
// RIGHT:
const opts = { recursive: true }
fs.mkdirSync(dir, opts)

Apply to ALL object literals: fs.mkdirSync, page.screenshot, page.waitForFunction, etc.


Phase 2: Implement Fix

Write minimum code to pass. Do NOT modify test assertions.

Debugging loop (most tests need 2-4 iterations):

  1. Read error carefully — 90% are selector/fixture mismatches, not logic bugs
  2. Verify DOM — use page.waitForSelector or page.waitForFunction to confirm elements exist
  3. Check prerequisites — hover before clicking, expand before asserting, wait for idle
  4. Use page.pause() to inspect live DOM
  5. Never weaken assertions — fix the code, not the test

Common failure patterns (see references/debugging-guide.md):

SymptomLikely CauseFix
Timeout on selectorWrong data attribute or shadow DOMCheck DOM with page.pause(), try ARIA roles
waitMark never resolvesSeed function missing expected contentMatch seed format exactly from passing tests
Button not visibleMissing hover/expand stepHover parent row, expand section first
Review panel emptyWrong changes mode (git vs session)Switch mode before asserting
Stale element referenceRace condition with async updatesAdd waitSessionIdle after mutations

Phase 3: AFTER Screenshot + Hard Gate

  await expect(fixedElement).toBeVisible()
  await screenshot.captureAfter("feature-working")

Hard gate — ALL three must be true before PR:

  • Test passes (exit code 0)
  • BEFORE-*.png exists
  • AFTER-*.png exists

Phase 4: PR + Screenshot Comment

# Stage, commit, push
git add -A
git commit -m "fix: <issue-description> (closes #<issue-number>)"
git push origin fix/<slug>

# Create PR referencing the issue
gh pr create \
  --title "fix: <short-description>" \
  --body "Closes #<issue-number>

## Summary
<Brief description of the fix>

## Screenshots
| Before | After |
|--------|-------|
| ![Before](<BEFORE-screenshot-url>) | ![After](<AFTER-screenshot-url>) |

## Verification
- [x] E2E test passes
- [x] BEFORE screenshot captured
- [x] AFTER screenshot captured" \
  --repo <owner/repo>

# Push screenshots to branch for GitHub raw URLs
git add e2e-screenshots/
git commit -m "chore: add e2e screenshots"
git push origin fix/<slug>

gh pr comment has no --attach — push images to branch, use raw GitHub URLs.


Backend-Only PRs

For PRs without direct UI changes, find the closest UI-exercisable path:

Change TypeE2E StrategyReal Example
Config utilityOpen settings dialogopenSettings(page) → screenshot config panel
Python resolverRun terminal commandrunTerminal(page, {cmd: "python3 --version"}) → show output
Race conditionTrigger rapid operationsCreate file externally → click refresh → verify
State managementMulti-session switchingSelect file in session A → switch to B → back to A
Cache bypassForce refreshAdd external file → click refresh button → verify
URL parsingCheck connection statusopenStatusPopover(page) → verify connected
Focus managementVerify focus after actionOpen file → check prompt still focused

NEVER use page.setContent() — always test through the real application.


Quick Reference

# Select an issue (manual or random)
gh issue list --repo owner/repo --state open --json number,title

# Create worktree
git worktree add .worktrees/fix/ISSUE -b fix/ISSUE
cd .worktrees/fix/ISSUE

# Set screenshot dir
export E2E_SCREENSHOT_DIR="$(pwd)/e2e-screenshots"

# Run E2E test
npm run test:e2e -- <test-file>

# Push and create PR
git push origin fix/ISSUE
gh pr create --title "fix: ..." --body "Closes #N" --repo owner/repo

Constraints

  • One issue at a time — complete one PR before starting the next
  • BEFORE and AFTER must BOTH exist before PR comment — hard gate, no exceptions
  • Screenshots save to worktree via E2E_SCREENSHOT_DIR, never to main repo
  • Always study ≥2 passing tests first — this is the #1 rule
  • NEVER use page.setContent() — real app only
  • Playwright TS transform rejects inline object params — always extract to variable
  • screenshot-reporter.ts must be copied to each worktree's e2e/ (not on fix branches)
  • Test MUST pass before PR — never create PR with failing tests

平台分布

OpenClaw

83.27%
按下载量换算755

安全审计

VirusTotal

未展示

ClawScan

可疑

Static analysis

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills