Token导航 LogoToken导航

Agent Skill · 研究检索

candid-validate-standards(坦诚验证标准)

candid-validate-standards 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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。

candid-validate-standards 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 41,842。

总安装 1,129周安装 48GitHub Stars 17下载 396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

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:candid-validate-standards(坦诚验证标准)
来源仓库:https://github.com/ron-myers/candid
仓库路径:skills/candid-validate-standards
安装命令:
npx skills add https://github.com/ron-myers/candid --skill candid-validate-standards
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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/ron-myers/candid --skill candid-validate-standards

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。

用于查找、检索和筛选相关信息,适合快速定位候选结果。适用于 Codex、Claude、Cursor、Gemini CLI 中的关键词搜索任务。

安装命令:npx skills add https://github.com/ron-myers/candid --skill candid-validate-standards 可结合来源仓库和原始 README 继续核验具体用法。建议确认权限范围和维护状态,避免触发联网或文件读写操作。

来源仓库内路径:skills/candid-validate-standards

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 Codex、Claude、Cursor、Gemini CLI;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 npx skills add https://github.com/ron-myers/candid --skill candid-validate-stand…;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

candid-validate-standards 怎么安装?

在终端执行 npx skills add https://github.com/ron-myers/candid --skill candid-validate-standards。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

candid-validate-standards 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。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。

candid-validate-standards 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

candid-validate-standards 是免费的吗?

来源标注的许可证是 MIT。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。

candid-validate-standards 用的人多吗?

公开数据是累计安装 1,129 次,近一周安装 48 次,下载 396 次,GitHub 17 星。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。

candid-validate-standards 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 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 Codex,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。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

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
candid-validate-standards

坦诚验证标准 · 本页

1,12917只读MIT通过

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

Technical.md Validator

Analyze a Technical.md file and identify rules that may be ineffective, vague, or redundant with existing tooling.

Workflow

Step 1: Locate Technical.md

Find the Technical.md file to validate:

  1. If path argument provided → use that path
  2. Check ./Technical.md
  3. Check ./.candid/Technical.md

If no file found:

❌ No Technical.md found.

Looked in:
- ./Technical.md
- ./.candid/Technical.md

Create one with: /candid-init

Step 2: Detect Linter Configs

Check for existing linter configurations:

# JavaScript/TypeScript
ls .eslintrc* eslint.config.* .prettierrc* biome.json 2>/dev/null

# Python
ls .flake8 pyproject.toml setup.cfg .ruff.toml ruff.toml 2>/dev/null

# Go
ls .golangci.yml .golangci.yaml 2>/dev/null

# Ruby
ls .rubocop.yml 2>/dev/null

Store detected linters for Step 4.

Step 3: Parse Technical.md

Read the Technical.md file and extract rules.

Rule detection:

  • Lines starting with - or * (list items)
  • Lines starting with numbers (numbered lists)
  • Lines following a ## heading

Skip:

  • Code blocks (between ```)
  • Empty lines
  • Comments (HTML or markdown)
  • Heading lines themselves

Store each rule with:

  • Line number
  • Section heading it belongs to
  • Full rule text

Step 4: Validate Each Rule

Check each rule against these issue categories:

4.1 Vague Language (🌫️)

Flag rules containing vague terms without specifics:

Vague TermWhy It's Vague
"clean"Subjective, no definition
"good"Subjective
"proper" / "appropriate"Undefined standard
"well-designed" / "well-structured"No criteria
"readable" / "maintainable"Without metrics
"best practices"Circular reference
"when necessary" / "when appropriate"Undefined trigger
"avoid" (alone)No guidance on alternatives
"consider"Not a requirement

Exception: Terms are OK if followed by specific criteria:

  • "readable (functions under 50 lines)" ✓
  • "maintainable" ✗
4.2 Missing Thresholds (📏)

Flag rules that imply quantity without numbers:

PatternIssue
"small functions"How small?
"short methods"How short?
"limit parameters"To how many?
"minimal dependencies"What's minimal?
"few levels of nesting"How few?
"reasonable timeout"What's reasonable?

Fix pattern: Add specific numbers (e.g., "functions under 50 lines")

4.3 Linter Overlap (🔧)

Flag rules that linters typically handle (based on Step 2):

If ESLint/Prettier detected:

  • Semicolon usage
  • Quote style (single vs double)
  • Indentation rules
  • Trailing commas
  • Import ordering
  • Unused variables
  • No console.log

If Flake8/Ruff detected:

  • Line length
  • Import sorting
  • Whitespace rules
  • Naming conventions (snake_case)

If any linter detected:

  • Formatting rules in general
  • Basic syntax style
4.4 Multiple Concerns (🎯)

Flag rules that try to do too much:

  • Rules with "and" connecting unrelated concerns
  • Rules over 2 lines long
  • Rules with multiple "must" statements

Example:

❌ "Functions should be small, well-documented, and handle errors properly"
✓ Split into three rules
4.5 Unverifiable Rules (❓)

Flag rules that can't be checked by reading code:

  • "Think about edge cases"
  • "Be consistent"
  • "Follow team conventions"
  • "Use common sense"
  • References to external documents without specifics

Step 5: Generate Report

Present findings organized by severity:

# Technical.md Validation Report

**File:** ./Technical.md
**Rules analyzed:** [N]
**Issues found:** [M]

---

## 🔴 Critical Issues

Rules that won't be effective in reviews.

### 🌫️ Vague Language

| Line | Rule | Issue |
|------|------|-------|
| 12 | "Write clean code" | "clean" is subjective |
| 24 | "Use appropriate error handling" | "appropriate" undefined |

### 📏 Missing Thresholds

| Line | Rule | Issue |
|------|------|-------|
| 18 | "Keep functions small" | No size specified |
| 31 | "Limit nesting depth" | No depth specified |

---

## 🟡 Warnings

Rules that may have issues.

### 🔧 Linter Overlap

| Line | Rule | Linter |
|------|------|--------|
| 8 | "Use semicolons" | ESLint handles this |
| 15 | "Sort imports alphabetically" | Prettier/ESLint handles this |

### 🎯 Multiple Concerns

| Line | Rule | Suggestion |
|------|------|------------|
| 22 | "Functions should be pure, small, and documented" | Split into 3 rules |

---

## ✅ Good Rules

[List of rules that passed validation]

Step 6: Suggest Fixes (if --fix flag)

If --fix argument provided, include specific rewrites:

## 💡 Suggested Rewrites

### Line 12: "Write clean code"
**Original:** Write clean code
**Suggested:**
- Functions must be under 50 lines
- No single-letter variable names except loop counters
- Maximum 3 levels of nesting

### Line 18: "Keep functions small"
**Original:** Keep functions small
**Suggested:** Functions must be under 50 lines (warning at 30)

### Line 24: "Use appropriate error handling"
**Original:** Use appropriate error handling
**Suggested:**
- All async functions must have try/catch or .catch()
- Errors must be logged with context before re-throwing
- User-facing errors must not expose stack traces

Step 7: Summary

End with actionable summary:

---

## Summary

- **[X] rules** are effective and specific ✅
- **[Y] rules** need thresholds or specifics 📏
- **[Z] rules** overlap with linters 🔧
- **[W] rules** are too vague to enforce 🌫️

### Recommended Actions

1. Remove [Z] linter-overlap rules (your linter handles these)
2. Add numbers to [Y] threshold rules
3. Rewrite [W] vague rules with specific criteria

Run `/candid-validate-standards --fix` for suggested rewrites.

Output Examples

Clean Technical.md

✅ Technical.md Validation Passed

File: ./Technical.md
Rules analyzed: 24
Issues found: 0

All rules are specific and verifiable. Nice work!

Issues Found

⚠️ Technical.md Validation: 8 issues found

File: ./Technical.md
Rules analyzed: 24
Issues found: 8

🌫️ Vague Language (3)
  Line 12: "clean code" - subjective term
  Line 18: "proper error handling" - "proper" undefined
  Line 31: "when necessary" - undefined trigger

📏 Missing Thresholds (2)
  Line 15: "small functions" - no size specified
  Line 22: "limit nesting" - no depth specified

🔧 Linter Overlap (3)
  Line 5: semicolons - handled by ESLint
  Line 8: quote style - handled by Prettier
  Line 11: import order - handled by ESLint

Run `/candid-validate-standards --fix` for suggested rewrites.

Vague Terms Reference

Use this list to detect vague language:

clean, good, proper, appropriate, suitable, adequate
well-designed, well-structured, well-organized, well-written
readable, maintainable, scalable, flexible, robust
best practices, industry standards, conventions
simple, straightforward, intuitive, obvious
reasonable, sensible, meaningful, significant
when necessary, when appropriate, when needed, as needed
avoid, prefer, consider, try to, should (without specifics)
minimal, few, some, many, several, various

Threshold Patterns Reference

Patterns that need numbers:

small/short/brief + (function|method|class|file|module)
limit/restrict/cap + (parameters|arguments|nesting|depth|complexity)
maximum/minimum + (without number following)
too many/too few + (without threshold)
keep ... under/below + (without number)
no more than + (without number)

Remember

The goal is to help users write Technical.md files that:

  1. Candid can actually enforce
  2. Don't duplicate existing tooling
  3. Focus on what matters (architecture, security, patterns)
  4. Are specific enough to be useful

A validated Technical.md leads to better code reviews.

平台分布

Claude Code

29.49%
按下载量换算117

Cursor

22.77%
按下载量换算90

Gemini CLI

18.48%
按下载量换算73

Codex

12.49%
按下载量换算49

windsurf

6.8%
按下载量换算27

OpenCode

3%
按下载量换算12

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。

继续浏览同类 Skills