claude-code

The Claudish Integration Skill provides a standardized, dynamic framework for Claude Code agents to interact with the Claudish CLI. By treating Claudish as the single source of truth for OpenRouter model data, this skill ensures that agents always have access to the latest pricing, context window specifications, and model categories without the need for hardcoded, static lists. It is designed for engineers and agent developers who require precision in model selection for complex coding workflows, automated code reviews, or multi-model validation pipelines.

• Enables real-time fetching of recommended model lists including pricing (input/output/average cost) and context size.

• Supports dynamic filtering by category such as coding, reasoning, vision, and budget-conscious models.

• Implements graceful fallback patterns to embedded defaults when Claudish is unavailable or unreachable.

• Allows for user-defined model overrides through CLAUDE.md configurations, prioritizing project-specific requirements.

• Facilitates JSON-based model parsing for seamless integration into command-line interface UIs and decision-making logic.

• Intended for use in commands requiring external AI model selection, proxy-mode agent delegation, and model metadata lookups.

• Requires local installation of the Claudish CLI (npm install -g claudish) to function as intended.

• Operates by executing shell commands (Bash) and parsing output to structure model objects.

• Should not be invoked when restricted strictly to embedded Claude models (Sonnet, Opus, Haiku) or in environments where external model API access is prohibited.

• Recommended for implementing high-velocity coding workflows that benefit from switching between specialized models like Grok, MiniMax, or GPT-5.