What problem does it solve? Claude's context window cannot hold very large codebases, making deep multi-file analysis, security audits, and architecture reviews impractical. This Skill routes those tasks to Gemini CLI, which offers a 1M token context window, and structures the delegation through subagents for iteration and verification. ## Core Features & Use Cases - Non-interactive Gemini queries: Run one-shot prompts with model selection (pro, flash, flash-lite aliases or full model IDs), output formats (text, JSON, stream-json), and directory scoping via --include-directories. - Subagent delegation pattern: Always dispatch Gemini work through the Task tool so the subagent can retry, request evidence from Gemini, and verify findings before summarizing. - Prompt pattern library: A references file provides ready-made templates for architecture reviews, security audits, performance analysis, migration assessments, and second-opinion critiques. - Use Case: You need a security audit of a 500k-token monorepo. Delegate to a subagent that runs Gemini with the security audit prompt template, collects evidence-backed findings, and returns a verified bullet-point summary. ## Quick Start Ask the agent to use the gemini-cli skill through a subagent to analyze your project directory for security vulnerabilities and summarize the verified findings.
