Menu

#869 by GitHub stars

google-gemma/gemma-skills

Skills for the Gemma and model/agent interactions

Engineering

History

Review weekly changes to SKILL.md before updating. Dates mark the weeks in which we observed each version; checks apply to the recorded commits.

View weekly changes

skills/gemma-dev/SKILL.md

Week of Nov 2, 2026 · +0 / −0 lines

Before Network or broad shell → After Network or broad shell

5aee63beeed3 → 5aee63beeed3

Compare on GitHub

Week of Oct 26, 2026 · Line counts unavailable

Before Not checked yet → After Network or broad shell

First observation → 5aee63beeed3

No earlier version was observed for comparison.

Compare on GitHub

skills/gemma-trainer/SKILL.md

Week of Nov 2, 2026 · +0 / −0 lines

Before Network or broad shell → After Network or broad shell

5aee63beeed3 → 5aee63beeed3

Compare on GitHub

Week of Oct 26, 2026 · Line counts unavailable

Before Not checked yet → After Network or broad shell

First observation → 5aee63beeed3

No earlier version was observed for comparison.

Compare on GitHub

View on GitHubAdd to my skills

Verified commit 5aee63beeed3d5591f7c5393519d7af1026b8a5c

1kGitHub stars
-1star in 7 days
2skills in this repository
4commits in 30 days

Heat

1.2
  • Star growth, last 7 daysAhead of 6% of tracked repositories
  • Star growth, last 24 hoursAhead of 3% of tracked repositories
  • Rank movement, last 7 daysAhead of 4% of tracked repositories
  • Commits, last 30 daysAhead of 67% of tracked repositories
  • Accounts asking about it through MCP, last 7 daysNot available
  • Hot votes this weekAhead of 0% of tracked repositories

Heat never changes the order by stars, and sponsorship never touches it. How heat works

Stars over time

Aug 2, 2026Nov 6, 20261,039

Individual skills in this repository

How it works

gemma-dev

When your agent uses it

Trigger this skill when building applications with Gemma or for general knowledge inquiries related to Gemma models (e.g. prompt structure, capabilities). Covers model selection, development workflows, and deployment best practices.

What it loads into context

Frontmatter

48 tokens

Always available to the agent

SKILL.md

2,517 tokens

When the skill is activated

Reference documents

0 tokens

Loaded on request

Token counts are estimates. Scripts run separately; they are not counted as reference context.

The playbook

  1. 1. Core Principle: Prioritize App Tooling
  2. 2. Model Selection Guide
  3. 3. Deployment Workflows
  4. 4. Speed Up Inference with Multi-Token Prediction (MTP)
  5. 5. Quantization-Aware Training (QAT)
  6. 6. Documentation Lookup

Scripts executed separately

  • assets/gradio-app.py
  • assets/transformers-js-app.js
  • assets/vertex-ai-app.py

Files

  • SKILL.mdcore9,448 bytes
  • SKILL.mdcore9,448 bytes
  • assets/gradio-app.pyruns1,977 bytes
  • assets/transformers-js-app.jsruns2,277 bytes
  • assets/vertex-ai-app.pyruns1,716 bytes
Read SKILL.md

# Gemma Development Skill

## 1. Core Principle: Prioritize App Tooling

**DO NOT** generate raw PyTorch, TensorFlow, or `transformers` code unless the
user explicitly asks for "Training," "Fine-tuning," or "Research." Always
default to high-level frameworks, SDKs, and tooling optimized for application
development.

## 2. Model Selection Guide

**CRITICAL:** Do not blindly default to `gemma-3-1b-it`. You must analyze the
user's specific domain, technical constraints, and required input modalities to
recommend the exact right fit. When recommending standard models, strictly
View the complete source on GitHub

Skills included

  • gemma-dev
  • gemma-trainer

Repository facts

License
Apache-2.0
Created
Apr 1, 2026
Last pushed
Oct 28, 2026
Forks
85
Open issues
5
Project activity
Active

Rank over the last 12 weeks

Now #869 · best #584 · down 130 in 30 days

  1. #850 on Nov 1
  2. #853 on Nov 2
  3. #860 on Nov 3
  4. #863 on Nov 4
  5. #868 on Nov 5
  6. #869 since Nov 6
55 earlier periods
  1. #595 on Aug 2
  2. #573 on Aug 7
  3. #561 on Aug 8
  4. #565 on Aug 9
  5. #591 on Aug 10
  6. #595 on Aug 11
  7. #590 on Aug 12
  8. #601 on Aug 13
  9. #597 on Aug 14
  10. #579 on Aug 15
  11. #592 on Aug 16
  12. #616 on Aug 18
  13. #604 on Aug 19
  14. #603 on Aug 20
  15. #608 on Aug 21
  16. #584 on Aug 22
  17. #612 on Aug 23
  18. #620 on Aug 24
  19. #629 on Aug 25
  20. #623 on Aug 27
  21. #641 on Aug 28
  22. #599 on Aug 29
  23. #628 from Sep 2 to Sep 3
  24. #609 on Sep 4
  25. #602 on Sep 5
  26. #614 on Sep 6
  27. #642 on Sep 10
  28. #635 on Sep 12
  29. #660 on Sep 16
  30. #647 on Sep 20
  31. #666 on Sep 21
  32. #677 on Sep 22
  33. #634 on Sep 26
  34. #664 on Sep 27
  35. #713 on Sep 29
  36. #654 on Oct 1
  37. #663 on Oct 2
  38. #682 on Oct 5
  39. #688 on Oct 6
  40. #698 on Oct 7
  41. #739 on Oct 8
  42. #710 on Oct 9
  43. #691 on Oct 11
  44. #712 on Oct 13
  45. #723 on Oct 14
  46. #721 on Oct 16
  47. #699 on Oct 17
  48. #730 on Oct 19
  49. #733 on Oct 23
  50. #698 on Oct 24
  51. #829 on Oct 25
  52. #837 on Oct 28
  53. #841 on Oct 29
  54. #844 on Oct 30
  55. #849 on Oct 31

Before you install

Network or broad shell

Automated signals, not a guarantee

Checked Nov 4, 2026

Not fully analyzed: some files were too long, unreadable or missing. This never lowers the level.

Show 2 notes
  • skills/gemma-dev/SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
  • skills/gemma-trainer/SKILL.md:1 Analysis incomplete: Referenced executable target unavailable; not analyzed
Show all 58 signals by file

skills/gemma-dev/SKILL.md

  • Network accessline 70
    -   This dedicated embedder supports up to 8k tokens with flexible output         dimensions (128 to 768). Fetch         [EmbeddingGemma 2 model card](https://gemma.ai/models/embeddinggemma/model-card-v2.md.txt)
  • Network accessline 71
    dimensions (128 to 768). Fetch         [EmbeddingGemma 2 model card](https://gemma.ai/models/embeddinggemma/model-card-v2.md.txt)         for the best practice.
  • Network accessline 72
    [EmbeddingGemma 2 model card](https://gemma.ai/models/embeddinggemma/model-card-v2.md.txt)
  • Network accessline 76
    -   This classifier is designed to run concurrently with your primary LLM to         ensure safety compliance. Fetch         [ShieldGemma 2 model card](https://gemma.ai/models/shieldgemma/model-card-v2.md.txt)
  • Network accessline 77
    ensure safety compliance. Fetch         [ShieldGemma 2 model card](https://gemma.ai/models/shieldgemma/model-card-v2.md.txt)         for the best practice.
  • Network accessline 78
    [ShieldGemma 2 model card](https://gemma.ai/models/shieldgemma/model-card-v2.md.txt)
  • Network accessline 98
    hardware. These configurations are optimized primarily for **Text**     workflows.     -   Easy Path: Use [Ollama](https://ollama.com/library/gemma4/tags) with
  • Network accessline 99
    workflows.     -   Easy Path: Use [Ollama](https://ollama.com/library/gemma4/tags) with         model tags like `gemma4:26b` or `gemma4:31b`.
  • Network accessline 100
    -   Easy Path: Use [Ollama](https://ollama.com/library/gemma4/tags) with
  • Network accessline 104
    `mlx_lm.lora`.     -   For more details, follow the         [MLX integration guide](https://gemma.ai/deployment/mlx.md.txt).
  • Network accessline 106
    [MLX integration guide](https://gemma.ai/deployment/mlx.md.txt).
  • Network accessline 129
    Fetch [Speed-up Gemma 4 with Multi-Token Prediction](https://gemma.ai/get-started/capabilities/mtp.md.txt)
  • Network accessline 130
    [Speed-up Gemma 4 with Multi-Token Prediction](https://gemma.ai/get-started/capabilities/mtp.md.txt)
  • Network accessline 186
    1.  Fetch the Index URL (`https://gemma.ai/llms.txt`) to discover available
  • Network accessline 190
    -   [Gemma 4 Best practices](https://gemma.ai/get-started/best-practices.md.txt)
  • Network accessline 191
    -   [Gemma 4 Prompt Formatting](https://gemma.ai/get-started/prompt-formatting.md.txt)
  • Network accessline 192
    -   [Text generation](https://gemma.ai/get-started/capabilities/text.md.txt)
  • Network accessline 193
    -   [Function calling](https://gemma.ai/get-started/capabilities/function-calling.md.txt)
  • Network accessline 194
    -   [Image understanding](https://gemma.ai/get-started/capabilities/image.md.txt)
  • Network accessline 195
    -   [Audio understanding](https://gemma.ai/get-started/capabilities/audio.md.txt)
  • Network accessline 196
    -   [Thinking mode](https://gemma.ai/get-started/capabilities/thinking.md.txt)
  • Network accessline 197
    -   [Text Embeddings](https://gemma.ai/models/embeddinggemma/inference.md.txt)
  • Network accessline 198
    -   [Multimodal Embeddings](https://gemma.ai/models/embeddinggemma/multimodal.md.txt)
  • Network accessline 199
    -   [Speed-up Gemma 4 with Multi-Token Prediction](https://gemma.ai/get-started/capabilities/mtp.md.txt)
  • Network accessline 200
    -   [DiffusionGemma](https://gemma.ai/models/diffusiongemma/explained.md.txt)

skills/gemma-dev/assets/gradio-app.py

  • Executable scriptline 1
    Executable file

skills/gemma-dev/assets/transformers-js-app.js

  • Executable scriptline 1
    Executable file

skills/gemma-dev/assets/vertex-ai-app.py

  • Executable scriptline 1
    Executable file
  • Access to secretsline 8
    [Credential-related source omitted]
  • Access to secretsline 9
    [Credential-related source omitted]
  • Access to secretsline 10
    [Credential-related source omitted]

skills/gemma-trainer/SKILL.md

  • Network accessline 218
    Fetch [Saving to GGUF](https://unsloth.ai/docs/basics/inference-and-deployment/saving-to-gguf.md)
  • Network accessline 219
    [Saving to GGUF](https://unsloth.ai/docs/basics/inference-and-deployment/saving-to-gguf.md)
  • Network accessline 224
    If you did not use Unsloth, you can convert your merged Hugging Face model directory manually using [llama.cpp](https://github.com/ggml-org/llama.cpp).
  • Network accessline 225
    directory manually using [llama.cpp](https://github.com/ggml-org/llama.cpp).
  • Network accessline 230
    E4B** on mobile, web, and IoT hardware with hardware acceleration (CPU, GPU, NPU). Fetch [LiteRT-LM guide](https://developers.google.com/edge/litert-lm/models/gemma-4.md.txt)
  • Network accessline 231
    NPU). Fetch [LiteRT-LM guide](https://developers.google.com/edge/litert-lm/models/gemma-4.md.txt) for the best practice.
  • Network accessline 232
    [LiteRT-LM guide](https://developers.google.com/edge/litert-lm/models/gemma-4.md.txt)

skills/gemma-trainer/assets/dataset_prep.py

  • Executable scriptline 1
    #!/usr/bin/env python3

skills/gemma-trainer/assets/distill_dataset.py

  • Executable scriptline 1
    #!/usr/bin/env python3
  • Network accessline 16
    import argparse from collections.abc import Sequence import http.client
  • Network accessline 17
    from collections.abc import Sequence import http.client import importlib
  • Network accessline 18
    import http.client
  • Network accessline 205
    headers = {"Content-Type": "application/json"}   req = urllib.request.Request(
  • Network accessline 206
    req = urllib.request.Request(
  • Network accessline 213
    try:     with urllib.request.urlopen(req, timeout=120) as response:
  • Network accessline 214
    with urllib.request.urlopen(req, timeout=120) as response:
  • Network accessline 222
    except (       OSError,       http.client.HTTPException,
  • Network accessline 223
    OSError,       http.client.HTTPException,       ValueError,
  • Network accessline 224
    http.client.HTTPException,
  • Network accessline 577
    "--ollama-url",       type=str,       default="http://localhost:11434",
  • Network accessline 578
    type=str,       default="http://localhost:11434",       help="URL of the local Ollama service endpoint",
  • Network accessline 579
    default="http://localhost:11434",

skills/gemma-trainer/assets/dpo_train.py

  • Executable scriptline 1
    #!/usr/bin/env python3

skills/gemma-trainer/assets/reward_train.py

  • Executable scriptline 1
    #!/usr/bin/env python3
  • Network accessline 7
    NOTE: AutoModelForSequenceClassification doesn't support Gemma 4 yet.       https://github.com/huggingface/transformers/issues/45373
  • Network accessline 8
    https://github.com/huggingface/transformers/issues/45373

skills/gemma-trainer/assets/sft_train.py

  • Executable scriptline 1
    #!/usr/bin/env python3

How the signals fit together

Facts from a static read of the files at the pinned commit. They describe what the code does, not why.

Signal confidence: 58 facts

and 8 more

Works with

Agent compatibility, from the files in this repository
AgentStatusEvidenceInstall to
Claude CodeCompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.claude/skills/<skill-name>/
CodexCompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.agents/skills/<skill-name>/
CursorCompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.cursor/skills/<skill-name>/
Gemini CLICompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.gemini/skills/<skill-name>/
GitHub CopilotCompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.github/skills/<skill-name>/
OpenCodeCompatibleStandard SKILL.md skills/gemma-dev/SKILL.md.opencode/skills/<skill-name>/
  • Native The repository ships a manifest, folder or frontmatter field for this agent.
  • Compatible A standard SKILL.md that this agent can read.
  • Unknown No evidence either way.

Derived automatically from paths, manifests and frontmatter. Not tested on each agent.

You might also use

StrinGhost/gemma-skills

Gemma 4 Agent Skills

54GitHub stars

obra/superpowers

An agentic skills framework & software development methodology that works.

313.6kGitHub stars

addyosmani/agent-skills

Production-grade engineering skills for AI coding agents.

113.7kGitHub stars

kepano/obsidian-skills

Agent skills for Obsidian. Teach your agent to use Obsidian CLI and open formats including Markdown, Bases, JSON Canvas.

50.9kGitHub stars

anthropics/claude-plugins-official

Official, Anthropic-managed directory of high quality Claude Code Plugins.

46.3kGitHub stars

mksglu/context-mode

Free, local plugin that keeps tool output out of your coding agent's context window. Sandbox and SQLite FTS5 search, 99% fewer bytes in context. Works with Claude Code, Codex, Pi,…

33.6kGitHub stars

Explore alternatives and comparisons

Add this badge to your README

It updates every night and links to this page.

google-gemma/gemma-skills: #869 on hotskills, 1k stars

[![google-gemma/gemma-skills on hotskills](https://hotskills.dev/badge/google-gemma-gemma-skills.svg)](https://hotskills.dev/skill/google-gemma-gemma-skills/)