Agent审计
- 作者仓库星标 206
- 作者更新于 实时读取
- 作者仓库 claude-ai-music-skills
- 领域
- 通用
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @bitwize-music-studio · 未声明 license
- 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
- 允许外网请求
- 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-model-updater
description: Updates model references across all skill files when new Claude models are released. Use when An…
category: 通用
runtime: 无特殊运行时
---
# skill-model-updater 输出预览
## PART A: 任务判断
- 适用问题：通用任务拆解、检查和交付。
- 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“Your Task / Step 1: Discover Current Models / Discovery Method”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通用任务拆解、检查和交付，并标出输入、限制和预期结果。
- **02** 执行计划：优先按“Your Task / Step 1: Discover Current Models / Discovery Method”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、会按任务需要访问外部网络、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；会按任务需要访问外部网络；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。 先确认触发方式
原文出现了 `/skill-model-updater` 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。
给清楚输入和边界
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。
小样例验证后再放大
先用一个小任务确认它会围绕“Your Task / Step 1: Discover Current Models / Discovery Method”工作；涉及文件或命令时，先看 diff、日志、预览或测试结果。
复核后再交付
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-model-updater
description: Updates model references across all skill files when new Claude models are released. Use when An…
category: 通用
source: bitwize-music-studio/claude-ai-music-skills
---
# skill-model-updater
## 什么时候使用
- 把通用方向的常用动作沉淀成 Agent 可调用的技能 适合处理通用任务拆解、检查、交付和复盘，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、可继续迭代的步骤；通常不需要额外…
- 面向通用任务拆解、检查和交付，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「Your Task / Step 1: Discover Current Models / Discovery Method」组织步骤，不把推断写成作者事实。
- 读取文件、写入/修改文件；会按任务需要访问外部网络；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。 证据边界与执行链路
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-model-updater" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Your Task / Step 1: Discover Current Models / Discovery Method
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 | 会按任务需要访问外部网络
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
}Your Task
Command: $ARGUMENTS
Based on the command:
- check - Discover current models, scan all skills, report status
- update - Update outdated models to current versions
- update --dry-run - Show what would be updated without making changes
Skill Model Updater
You maintain model currency across all skill files, ensuring skills use the latest Claude models.
Step 1: Discover Current Models
Before checking or updating skills, you MUST first discover current model IDs.
Discovery Method
Search for current Anthropic models:
WebSearch: "Anthropic Claude model IDs 2025" OR "Claude API models list current"Fetch official documentation:
WebFetch: https://docs.anthropic.com/en/docs/about-claude/modelsExtract current model IDs for each tier:
- Opus (most capable) - Look for
claude-opus-*orclaude-*-opus-* - Sonnet (balanced) - Look for
claude-sonnet-*orclaude-*-sonnet-* - Haiku (fast) - Look for
claude-haiku-*orclaude-*-haiku-*
- Opus (most capable) - Look for
Identify the latest version of each tier by date suffix (e.g.,
20250514>20250114)
Expected Output Format
After discovery, report:
CURRENT CLAUDE MODELS (discovered)
==================================
Source: docs.anthropic.com/en/docs/about-claude/models
Date checked: [today's date]
Opus: claude-opus-4-5-20251101
Sonnet: claude-sonnet-4-5-20250929
Haiku: claude-haiku-4-5-20251001Shorthand aliases (always valid, resolve to current):
opus→ current opus modelsonnet→ current sonnet modelhaiku→ current haiku model
Workflow
Check Mode
/skill-model-updater check- Discover current models (Step 1 above) - WebSearch/WebFetch Anthropic docs
- Glob for all
skills/*/SKILL.mdfiles - Extract
model:field from YAML frontmatter - Compare against discovered current models
- Check CLAUDE.md - Scan
${CLAUDE_PLUGIN_ROOT}/CLAUDE.mdforCo-Authored-By: Claudelines and verify model name is current - Report status for each skill and CLAUDE.md
Output format:
SKILL MODEL AUDIT
=================
Current Models (discovered from docs.anthropic.com):
- Opus: claude-opus-4-5-20251101
- Sonnet: claude-sonnet-4-5-20250929
- Haiku: claude-haiku-4-5-20251001
Skill Status:
✓ lyric-writer: claude-opus-4-5-20251101 (current)
✓ researcher: claude-sonnet-4-5-20250929 (current)
⚠ album-art-director: claude-sonnet-4-20250114 (outdated → claude-sonnet-4-5-20250929)
✓ import-audio: claude-haiku-4-5-20251001 (current)
Summary: 19/20 skills current, 1 needs updateUpdate Mode
/skill-model-updater update- Discover current models (Step 1 above)
- Run check to identify outdated skills
- For each outdated skill:
- Read the SKILL.md file
- Update the
model:field to discovered current version - Preserve the skill's tier (don't change opus to sonnet)
- Update CLAUDE.md - If
Co-Authored-Byline in${CLAUDE_PLUGIN_ROOT}/CLAUDE.mdreferences an outdated model name, update it to current - Report changes made
Output format:
SKILL MODEL UPDATE
==================
Models discovered from docs.anthropic.com:
- Opus: claude-opus-4-5-20251101
- Sonnet: claude-sonnet-4-5-20250929
- Haiku: claude-haiku-4-5-20251001
Updated 1 skill:
- album-art-director: claude-sonnet-4-20250114 → claude-sonnet-4-5-20250929
All skills now current.Dry Run Mode
/skill-model-updater update --dry-runSame as update but only reports what would change without editing files.
Model Detection Logic
Identifying Outdated Models
A model is outdated if:
- It's an older version of a current model family (e.g.,
claude-sonnet-4-20250114vsclaude-sonnet-4-5-20250929) - It's a deprecated model (e.g.,
claude-3-opus-20240229)
Tier Detection (Auto)
Detect tier from the skill's existing model: field - no hardcoded tier list needed:
- If model contains
opus→ update to current opus - If model contains
sonnet→ update to current sonnet - If model contains
haiku→ update to current haiku - If model is shorthand (
opus,sonnet,haiku) → leave as-is (always resolves to current)
This preserves deliberate tier assignments without maintaining a separate mapping.
When New Models Release
This skill discovers models automatically and detects tiers from existing assignments.
When Anthropic releases new models:
- Run check -
/skill-model-updater checkwill discover new models automatically - Review changes - Verify discovered models are correct
- Run update -
/skill-model-updater updateto propagate changes
Note: Tier assignments are documented in ${CLAUDE_PLUGIN_ROOT}/reference/model-strategy.md. This skill preserves existing tiers - it only updates version numbers.
Example: Full Update Cycle
User: "New Claude models released, update skills"
1. Run check (discovers models automatically):
/skill-model-updater check
Output:
- Discovered from docs.anthropic.com: Opus 4.5, Sonnet 4, Haiku 3.5
- 3 skills using outdated sonnet (20250114 → 20250514)
- 1 skill using deprecated opus (claude-3-opus → claude-opus-4-5)
2. Run dry-run:
/skill-model-updater update --dry-run
Output shows proposed changes
3. Run update:
/skill-model-updater update
Output confirms 4 skills updated
4. Verify:
/skill-model-updater check
Output: All 21 skills currentError Handling
Missing Model Field
If a SKILL.md has no model: field:
- Report as "⚠ [skill]: No model specified"
- Do not add one automatically (requires manual decision)
Unknown Model
If a SKILL.md has an unrecognized model:
- Report as "? [skill]: Unknown model '[model-id]'"
- Do not update (requires manual review)
Invalid YAML
If a SKILL.md has malformed frontmatter:
- Report as "✗ [skill]: Invalid YAML frontmatter"
- Do not attempt to update
Scope
This skill updates model references in:
- All
skills/*/SKILL.mdfiles - Themodel:field in YAML frontmatter CLAUDE.md- TheCo-Authored-By: Claude [Model] <noreply@anthropic.com>line in the versioning section
Both locations must stay in sync with the latest Claude model names.
Remember
- Check before update - Always know what will change
- Tiers are auto-detected - Skill reads existing model field to determine tier (opus/sonnet/haiku)
- Shorthand is safe -
opus,sonnet,haikualways resolve to current versions - Tier rationale - See
${CLAUDE_PLUGIN_ROOT}/reference/model-strategy.mdfor why each skill uses its tier - CLAUDE.md co-author line - Must reflect the current top-tier model name used for commits
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