github-search-before-code

Proactively search GitHub for reference implementations before writing new code. Use this skill when: (1) User requests implementing completely new functionality, algorithms, or modules that don't exist in the current codebase, (2) User mentions repeated failures with phrases like 'still not working', 'tried many times', 'still has problems', or (3) AI recognizes the need to implement unfamiliar or complex features. The skill helps avoid reinventing the wheel by finding and analyzing existing high-quality implementations, then adapting them to user needs.

GitHub Search Before Code

Trigger Conditions

Scenario 1 - New functionality: New algorithms, protocols, drivers, or complex features not in codebase Scenario 2 - Repeated failures: User says "still not working", "tried many times", "改了很多次了", "还是有问题"

Workflow

1. Infer Domain

Analyze conversation context to extract 1-2 domain keywords:

Without first identifying the industry background, keyword extraction may lead to results that are completely off-topic.
Current discussion topic
Recent code files/functions
User's project description

If unclear, see industry-background.md for domain keyword reference.

2. Construct Keywords

Pattern: "<function> <domain> <tech>"

Examples:

C: "harmonic analysis" + "power" → "harmonic analysis power metering" C
Python: "web scraping" → "web scraping beautifulsoup" Python
Shell: "backup automation" → "incremental backup script" Shell

Domain matters: "Goertzel" → audio ❌, "Goertzel power" → power analysis ✅

3. Search GitHub

python3 scripts/github_search.py repo "<keywords>" [language]

Features:

see search-strategies.md for search flow reference.
Auto-fallback: tries 4 strategies if no results (language → no-lang → simplified → core-word)
Token support: set GITHUB_TOKEN env var for 5000/h quota (vs 60/h)
Desc truncated to 200 chars to save tokens

4. Analyze & Use

Screen: Stars > 20, language matches project, updated < 2yr Read: WebFetch README for key info Use: Extract logic, adapt style, add // Reference: [URL] (or # Reference: for Python/Shell)

No results: Try broader keywords or WebSearch

Commands

Output: ⭐stars | lang | year | desc (sorted by stars)