Generate LLM outputs constrained by regex and grammar across multiple frameworks and workflows.
npx skills add https://github.com/ricable/mcai --skill guidance-ricable
Please help me install this Agent Skill. Skill: guidance Source: https://github.com/ricable/mcai/tree/main/.agents/skills/guidance Command: npx skills add https://github.com/ricable/mcai --skill guidance-ricable
Guidance removes the uncertainty of free-form model prompting by constraining LLM output so it matches the format, schema, or workflow you need. It helps prevent invalid JSON, malformed code, inconsistent classifications, and brittle retry loops.
Ask Guidance to generate a valid structured response, such as a JSON object or classified label, using constrained generation and the backend of your choice.
High-intent search queries and answers about installing and using this skill.
Constrained generation applies regex and grammar controls to force valid JSON, XML, CSV, or code from LLMs. This prevents malformed structures and eliminates brittle retry loops by ensuring exact format compliance without manual post-processing.
Invalid JSON occurs during free-form prompting because models lack structural constraints. Applying grammar-based generation forces the LLM to output valid JSON, eliminating parsing errors and removing the need for retry loops or manual extraction.
Constrained generation works across Anthropic, OpenAI, Transformers, and llama.cpp backends. It applies token healing and categorical selection controls specific to each backend to maintain format consistency across both API and local model deployments.
Workflow orchestration coordinates multi-step reasoning, ReAct-style tool use, and extraction pipelines to build agents. It enforces structured output at each step, ensuring validated data handoffs between tools without intermediate parsing failures.
Token healing is required when applying constrained generation across different LLM backends. It corrects token boundary mismatches to ensure grammar and regex constraints produce exact categorical selections and valid structured output.
musical-basicsConstrain LLM output with regex and grammar rules for valid JSON, XML, and code.
peteromalletConstrain LLM outputs with regex and grammars using Guidance.
cxnaiveConstrain LLM output syntax with regex, grammars, and token-level filtering.
carterwayneskhizeineGenerate regex- or grammar-constrained JSON, XML, and code outputs.
