AgentSkillsCN

clone-analyzer

对于任何GitHub仓库，建议采用浅克隆方式在本地进行深度代码分析，而非依赖Web API。

SKILL.md
--- frontmatter
name: clone-analyzer
description: For any GitHub repository, prefer cloning locally with shallow clone for deep code analysis using local file tools instead of web APIs
license: MIT
compatibility: opencode
metadata:
  audience: developers
  workflow: analysis

What I do

When a GitHub repository is referenced (URL, org/repo format, or by name), I clone it locally for thorough analysis using local file tools (Read, Grep, Glob, AST-grep, LSP) rather than relying on web-based browsing or API calls.

When to use me

  • •A GitHub repository URL or project reference appears and code analysis is needed
  • •You need to understand how an external project works, find patterns, or review implementations
  • •Web-based tools give incomplete or truncated results

Instructions

  1. •

    Clone to a temp directory using shallow clone to minimize download:

    bash
    git clone --depth 1 <repo-url> /tmp/<repo-name>

    For very large repos, use sparse checkout or blob filter:

    bash
    git clone --depth 1 --filter=blob:none <repo-url> /tmp/<repo-name>
  2. •

    Analyze locally using file tools:

    • •Read to examine files
    • •Grep / AST-grep to search patterns
    • •Glob to discover file structure
    • •LSP tools if the language server supports it
  3. •

    Reuse existing clones — if /tmp/<repo-name> already exists, don't re-clone

  4. •

    Clean up when analysis is complete and the clone is no longer needed:

    bash
    rm -rf /tmp/<repo-name>

Key rules

  • •Always use --depth 1 to keep clones fast and small
  • •Clone to /tmp/ to avoid polluting the workspace
  • •Prefer local clone analysis over gh api, web fetch, or grep.app for comprehensive understanding
  • •If only a small piece of info is needed (e.g., a single file), web tools may suffice — use judgment