Skip to content

GEMINI.md

Latest commit

 

History

History
94 lines (68 loc) · 5.36 KB

GEMINI.md

File metadata and controls

94 lines (68 loc) · 5.36 KB
Raw

Claude-Agents: Skills Ecosystem for Gemini CLI

You have access to 153 specialized skills, 185 specialized agents, and 100 commands organized across 80 plugins. These skills are designed for progressive disclosure — they activate automatically when you describe a matching task, saving context tokens.

Navigation

  • Skill Discovery: Describe your task (e.g., "Set up a Kubernetes deployment") and Gemini CLI will identify the matching skill and ask to activate it.
  • Opt-In Slash Commands: Slash commands are optional and not available until you generate them locally. This keeps your namespace clean. See Setup below.
  • Plugin Catalog: See docs/plugins.md for the full catalog of all 80 plugins and 153 skills.

Safety and System Integrity

  • Credential Protection: Agents will never log, print, or commit secrets, API keys, or sensitive credentials.
  • Source Control: Changes are never staged or committed unless specifically requested by the user.
  • Transparent Execution: Complex workflows pause at critical checkpoints for user approval.
  • Isolation: Work is confined to the project directory; system-level configurations are protected.

Setup: Opt-In Slash Commands

Slash commands are generated on-demand to avoid cluttering your CLI namespace.

  1. Install the extension: gemini extensions install https://github.com/wshobson/agents
  2. Generate commands for a plugin: 
    cd ~/.gemini/extensions/claude-code-workflows
make generate-plugin PLUGIN=javascript-typescript
    (Windows: run in Git Bash or WSL, or use python3 tools/generate_gemini_commands.py --plugin <name> directly.)
  3. Keep in sync after updates: 
    cd ~/.gemini/extensions/claude-code-workflows
make sync-commands
  4. Restart Gemini CLI for new commands to appear.

Interactive Execution (Protocol Orchestrator)

Slash commands follow a sequential, multi-step protocol model. When you run a command like /tdd-workflows:tdd-cycle, the agent:

  1. Reads the full protocol from the repository's source Markdown.
  2. Pauses at checkpoints for your approval (e.g., PHASE CHECKPOINT).
  3. Tracks progress locally, allowing you to resume or audit the workflow.

Skill Library (Grouped by Plugin)

Below are the primary skills available to you. Describe your goal to trigger them.

Language Development

  • python-development: async-python-patterns, python-packaging, python-testing-patterns, uv-package-manager
  • javascript-typescript: javascript-testing-patterns, modern-javascript-patterns, nodejs-backend-patterns, typescript-advanced-types
  • systems-programming: go-concurrency-patterns, memory-safety-patterns, rust-async-patterns
  • shell-scripting: bash-defensive-patterns, bats-testing-patterns, shellcheck-configuration

Full-Stack & UI

  • backend-development: api-design-principles, architecture-patterns, cqrs-implementation, event-store-design, saga-orchestration, temporal-python-testing
  • frontend-mobile-development: nextjs-app-router-patterns, react-native-architecture, react-state-management, tailwind-design-system
  • ui-design: accessibility-compliance, design-system-patterns, interaction-design, mobile-ios-design, responsive-design, visual-design-foundations

Infrastructure & Security

  • cloud-infrastructure: cost-optimization, istio-traffic-management, multi-cloud-architecture, terraform-module-library
  • kubernetes-operations: gitops-workflow, helm-chart-scaffolding, k8s-manifest-generator, k8s-security-policies
  • security-scanning: attack-tree-construction, sast-configuration, stride-analysis-patterns, threat-mitigation-mapping
  • reverse-engineering: anti-reversing-techniques, binary-analysis-patterns, memory-forensics, protocol-reverse-engineering

... and 50+ more. See docs/plugins.md for the full catalog.

Slash Commands

100 slash commands are available across 50 plugins, mirroring Claude Code's /plugin:command namespace:

/security-scanning:security-sast          # SAST vulnerability scan
/backend-development:feature-development  # End-to-end feature orchestration
/tdd-workflows:tdd-cycle                  # Full TDD red-green-refactor
/python-development:python-scaffold       # Python project scaffolding

Use /help in Gemini CLI to list all available commands.

Key Differences from Claude Code

Feature Claude Code Gemini CLI Workaround
Slash Commands Built-in /plan, /spec, /ship No built-in IDE commands Use opt-in extension commands (see Setup)
Subagent Orchestration Fan-out parallel execution Sequential execution Use the executing-plans skill to batch tasks
Model Assignment Per-agent model tiers (Opus/Sonnet/Haiku) Session-level model selection Skills are model-agnostic; use session default
Plugin Installation Per-plugin via /plugin install Per-extension via gemini extensions install Install once: gemini extensions install https://github.com/wshobson/agents

Support & Contribution

  • Issues: Report bugs or suggest skills at https://github.com/wshobson/agents/issues
  • Skill Development: Contribute new skills by creating a skills/<skill-name>/SKILL.md file
  • Plugin Development: Create new plugins in plugins/ following the structure in CLAUDE.md