Agent诊断
- 作者仓库星标 1,104
- 许可证 Complete terms in LICENSE.txt
- 作者更新于 实时读取
- 作者仓库 archive
- 领域
- 工程开发
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 94 / 100 · 已通过审计
- 作者 / 版本 / 许可
- @dp-archive · Complete terms in LICENSE.txt
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: skill-updater
description: Update existing skills based on feedback. Use when users want to improve or modify an existing s…
category: 工程开发
runtime: 无特殊运行时
---
# skill-updater 输出预览
## PART A: 任务判断
- 适用问题:代码实现、重构、调试或代码审查。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“When to Use / Required Inputs / Update Process”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于代码实现、重构、调试或代码审查,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“When to Use / Required Inputs / Update Process”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文出现了 `/tmp` 这类斜杠命令;如果你的 Agent 支持命令触发,优先用命令开场,再补充目标和边界。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件。
先用一个小任务确认它会围绕“When to Use / Required Inputs / Update Process”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: skill-updater
description: Update existing skills based on feedback. Use when users want to improve or modify an existing s…
category: 工程开发
source: dp-archive/archive
---
# skill-updater
## 什么时候使用
- 把工程方向的常用动作沉淀成 Agent 可调用的技能 适合处理工程开发场景下的代码实现、调试、重构、测试或代码审查,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代…
- 面向代码实现、重构、调试或代码审查,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「When to Use / Required Inputs / Update Process」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "skill-updater" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> When to Use / Required Inputs / Update Process
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Skill Updater
This skill provides guidance for updating existing skills based on user feedback.
When to Use
Use this skill when:
- User provides feedback about an existing skill's performance
- User wants to improve or modify a skill
- User reports issues or bugs with a skill
- User wants to add new functionality to an existing skill
Required Inputs
- Skill Path: Path to the skill directory or .skill file to update
- Feedback: Description of what needs to change, improve, or be fixed
Update Process
Step 1: Load and Understand the Skill
Use the analyze_skill.py script to get a quick overview:
scripts/analyze_skill.py <skill-path>
This shows:
- Skill name and description
- SKILL.md line/word count
- All bundled resources (scripts/, references/, assets/)
- Any structural issues
Then read SKILL.md to understand the detailed instructions:
cat <skill-path>/SKILL.md
For .skill files (zip archives), extract first:
unzip <skill-path>.skill -d /tmp/skill-extract/
Step 2: Analyze the Feedback
Categorize the feedback into one or more types:
| Type | Example Feedback | Typical Changes |
|---|---|---|
| Bug fix | "The script fails when..." | Fix scripts, add error handling |
| Missing feature | "It should also support..." | Add new functionality |
| Clarity | "Instructions are confusing" | Improve SKILL.md documentation |
| Performance | "Too slow/uses too much context" | Optimize scripts, reduce SKILL.md |
| Trigger | "Doesn't activate when I say..." | Update description in frontmatter |
Step 3: Plan the Changes
Before making changes, create a clear plan:
- List specific files to modify
- Describe what changes each file needs
- Identify any new files to create or files to delete
- Consider impact on other parts of the skill
Step 4: Implement Changes
Apply changes following skill design principles:
For SKILL.md changes:
- Keep body under 500 lines
- Use imperative form
- Move detailed content to references/ if approaching limit
- Ensure frontmatter description captures all trigger scenarios
For script changes:
- Test scripts after modification
- Maintain existing interfaces unless feedback requires changes
- Add error handling for reported edge cases
For reference changes:
- Keep references one level deep from SKILL.md
- Update SKILL.md to reference new files
Step 5: Validate and Package
After implementing changes:
# Validate the skill
scripts/quick_validate.py <skill-path>
# Package if validation passes
scripts/package_skill.py <skill-path>
Common Update Patterns
Pattern 1: Fixing Script Bugs
- Read the failing script
- Identify the bug based on feedback
- Apply fix
- Test with the scenario from feedback
- Package
Pattern 2: Improving Trigger Accuracy
- Review current frontmatter description
- Identify missing trigger phrases from feedback
- Update description to include new triggers
- Ensure description remains concise (<100 words)
Pattern 3: Adding New Functionality
- Determine if new functionality needs:
- New script in scripts/
- New reference in references/
- New asset in assets/
- Updates to SKILL.md instructions
- Implement required components
- Update SKILL.md to document new functionality
- Test end-to-end
Pattern 4: Reducing Context Usage
- Identify verbose sections in SKILL.md
- Move detailed content to references/
- Replace with concise summary + link
- Remove redundant examples
- Verify skill still functions correctly
Best Practices
- Minimal changes: Only change what the feedback requires
- Test after changes: Run scripts and validate skill structure
- Preserve intent: Maintain the skill's original purpose
- Document changes: Update SKILL.md if behavior changes
- Version awareness: If the skill has users, consider backward compatibility
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核