Archived Google Gemma Skill

This historical integration is retained for reference but is outside active agent-skill discovery. Codex, Copilot, and Claude use the canonical playbooks under .github/skills/.

About Google Gemma

Google Gemma is a family of lightweight open models built from the same research used for Gemini. In this repository, Gemma is configured as a SHAFT-focused expert for code diagnosis, issue fixing guidance, and test generation.

Skill Structure (Google AI Edge Gallery Format)

This directory is a self-contained AI Edge Gallery skill.

google-gemma/
├── SKILL.md                              ← Gallery-format skill definition
├── scripts/
│   └── index.html                        ← JS helper to fetch GitHub source files
├── overview.md                           <- This file
└── code-analysis-and-optimization.md     ← Extended analysis reference

Historical Use

This snapshot documents the former AI Edge Gallery integration. It is not a supported active agent workflow and should not be loaded as repository policy.

Available Skills

Skill	File	Description
SHAFT_ENGINE Expert	SKILL.md	High-accuracy SHAFT issue triage, root-cause analysis, and minimal-risk fix guidance
Code Analysis and Optimization (reference)	code-analysis-and-optimization.md	Broader performance/architecture analysis reference

Enhanced Capabilities

Capability	Mode	Value
Analyze pasted Java code	Text	Detects SHAFT anti-patterns with confidence scoring
Fetch and analyze GitHub file URLs	JS + Text	Converts supported GitHub URLs to raw source with validation and timeout handling
Prioritize fixes	Text	Produces risk-based fix order and verification steps
Generate SHAFT-compliant tests	Text	Produces TestNG/JUnit5 skeletons aligned to SHAFT lifecycle patterns
Improve issue-fix accuracy	Text	Distinguishes root cause vs symptoms and flags missing context

Notes

• Best results come from including setup/teardown + test methods + relevant stack trace.
• The helper script only supports public GitHub file URLs.
• The skill emphasizes minimal safe changes and explicit validation guidance.