reasoning-gaps

reasoning-gaps is a command for Claude Code from paulbaranowski/wild-horses. It costs 73 tokens per session (3,699 once invoked), scanned A, original, MIT.

A code analysis command that looks for places where an AI agent may misunderstand data flow or runtime behavior. It examines issues such as unclear types, hidden state, implicit control flow, missing documentation, and complex structure.

In plain words
What is it for?
Use it to find AI reasoning gaps in a codebase and create concrete tasks for improving clarity and structure.
Why use it?
Normal code-quality checks may pass while the code remains difficult for an AI agent to trace. This analysis produces a prioritized plan for making that reasoning easier.

Command for Claude Code

Written for Claude Code: argument-hint in frontmatter. Also seen: mentions CLAUDE.md; mentions subagents.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the harness plugin — 3 skills, 2 commands shipped together

Good fit Use it to find AI reasoning gaps in a codebase and create concrete tasks for improving clarity and structure.

Compare 6 commands from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add paulbaranowski/wild-horses
Claude Code
/plugin install harness

Made for: Claude Code.

Or install harness, the plugin that ships this one along with the rest of its 3 skills, 2 commands.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for reasoning-gaps

README.md
[![agentmods](https://agentmods.dev/badge/commands/paulbaranowski/wild-horses/reasoning-gaps/github.svg)](https://agentmods.dev/commands/paulbaranowski/wild-horses/reasoning-gaps)
Your own site
<a href="https://agentmods.dev/commands/paulbaranowski/wild-horses/reasoning-gaps"><img src="https://agentmods.dev/badge/commands/paulbaranowski/wild-horses/reasoning-gaps/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for reasoning-gaps

Your own site · 80×15
<a href="https://agentmods.dev/commands/paulbaranowski/wild-horses/reasoning-gaps"><img src="https://agentmods.dev/badge/commands/paulbaranowski/wild-horses/reasoning-gaps.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 73 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,699 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00073 $0.03699
Opus 5 $0.00036 $0.01850
Sonnet 5 $0.00015 $0.00740
Haiku 4.5 $0.00007 $0.00370

Measured 10d ago against content hash fb9a0d4facdf, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

reasoning-gaps scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/harness/commands/reasoning-gaps.md · 197 lines

How it starts

The opening of the file, as written. The whole thing — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AI Reasoning Gap Analysis

Analyze code for AI reasoning gaps — places where an AI agent cannot confidently trace data flow, predict runtime behavior, or orient itself in the codebase. Uses 3 parallel specialist agents, then synthesizes findings into a prioritized remediation plan with concrete interventions.

This is NOT a code quality review. Code can be well-written and still be opaque to AI reasoning. This skill answers: "If an AI agent read this code, what would it get wrong?"

Bundled assets at ${CLAUDE_PLUGIN_ROOT} (if the variable isn't substituted in this context, find the files with Glob "**/harness/loop-protocol.md" and read the siblings alongside it):

  • loop-protocol.md — Phase 4 options menu. Shared with /harness:feedback-blockers.
  • task-list-schema.md — JSON task file shape. Shared with /harness:feedback-blockers, task-list-builder, and task-list-runner.
  • skills/task-list-builder/SKILL.md — task-list construction (verifySteps discovery, run-ID, JSON + MD writing, preview). Invoked from Phase 4 Options 1, 2, and 3 with --slug reasoning-gaps --md-body-from-context.
  • skills/task-list-runner/SKILL.md — execution engine (resume, Agent loop, Task Implementation Prompt). Invoked from Phase 4 Options 1 and 2 and from --resume.
  • agents/reasoning-gaps/types-and-data-contracts.md
  • agents/reasoning-gaps/implicit-flow-and-state.md
  • agents/reasoning-gaps/structure-and-documentation.md

Target: "$ARGUMENTS"

Slug: reasoning-gaps (use this value wherever loop-protocol.md says <slug>).


Resume Check (before Phase 1)

If $ARGUMENTS contains --resume, hand off to the task-list-runner skill (it will auto-locate the in-progress task file or accept a path that follows --resume). Skip Phases 1–4 entirely.


Phase 1: Determine Scope (you do this)

Based on arguments and context, determine what files to analyze:

  1. If a specific file/directory path is given — collect those file paths.
  2. If a free-form description is given (e.g., "the cli code", "the decoder pipeline", "authentication logic") — search the codebase to identify matching files. Use directory names, module names, class/function names, and file contents to resolve the description to a concrete list of files. Confirm the resolved scope with the user if ambiguous.
  3. If no arguments (DEFAULT) — get only the files changed in the current PR branch: git diff --name-only main...HEAD plus any uncommitted changes via git diff --name-only. Exclude test files. This should typically yield 3–10 files. If it yields more than 15, ask the user to narrow scope.
  4. If --full — collect all source files in src/ or the main package directory (warn: may be slow).

Read the full file on GitHub · 197 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 10d ago First seen · 197 lines · 73 tokens per session scan A fb9a0d4facdf

Subscribe to this mod's changes

reasoning-gaps is a command published in the GitHub repository paulbaranowski/wild-horses (12 stars, last pushed 22d ago), licensed MIT. It adds 73 tokens to every session and 3,699 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.