reasoning-validator — Claude Code subagent: what it does
The reasoning-validator subagent for Claude Code checks reasoning chains with WFGY formulas and quantitative metrics, helping identify logical errors in assistant reasoning.
no license ★ 1,357 updated 2026-03-01
.claude/agents/reasoning-validator.mdWhat reasoning-validator does
This subagent adopts the Euclid persona on top of the Sonnet model, treating reasoning chains as mathematical structures. It tracks numerical indicators like semantic tension ΔS and reasoning resonance. If the resonance metric exceeds 0.3, the agent rejects the conclusion and triggers corrective formulas to reduce semantic residue.
Operation relies on internal commands prefixed with an asterisk. You can run *validate-chain to inspect reasoning steps, *apply-bbpf to explore parallel solution paths, or *metric-report to inspect calculated values. Upon launch, it runs initialization with /wfgy:init and attempts to pull formula parameters from .wfgy/config.json.
Execution depends on markdown command definitions located in .claude/commands/wfgy/ and .claude/commands/reasoning/. Without those files present in the project tree, external validation commands will fail to run.
What it helps with
- test complex deduction chains against quantitative thresholds
- detect fallacies and logic gaps using WFGY mathematical formulas
- measure semantic tension and stability scores during problem solving
- trace alternative deduction paths in parallel
Example prompts for the agent
- Check my reasoning steps using *validate-chain
- Calculate semantic tension across the argument with *calc-tension
- Explore alternative reasoning paths using *multi-path
How to install reasoning-validator
Manually: put the files in place
.claude/agents/reasoning-validator.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/reasoning-validator.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/reasoning-validator.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about reasoning-validator
How to install reasoning-validator in Claude Code?
Save the subagent configuration file directly into .claude/agents/reasoning-validator.md within your project root.
What does reasoning-validator do?
It assesses reasoning steps using WFGY equations, calculating numerical scores for semantic residue and resonance. When reasoning fails thresholds, it applies formulaic adjustments.
Which project files are required for operation?
The subagent looks for command files in .claude/commands/wfgy/ and .claude/commands/reasoning/. It also reads configuration parameters from .wfgy/config.json when present.