Verified:GitHub

hard-cheese

Grade author explanations of code diffs against a SOLO rubric.

15|1|Updated May 4, 2026

Search Tags:#code review #audit trail #causal reasoning #Socratic prompts #epistemic debt #fresh-context judge #SOLO taxonomy

Ready to run

npx skills add https://github.com/paulnsorensen/easy-cheese --skill hard-cheese

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: hard-cheese
Source: https://github.com/paulnsorensen/easy-cheese/tree/main/skills/hard-cheese
Command: npx skills add https://github.com/paulnsorensen/easy-cheese --skill hard-cheese

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It reduces epistemic debt by requiring a human author to explain the causal logic of a diff before code is shared for review or a pull request is opened.

Core Features & Use Cases

Metacognitive gate: Forces the author to articulate how and why the change produces the desired behavior, not just what changed.
Fresh-context judge grading: Spawns a separate judge sub-agent to evaluate the explanation against a SOLO-based rubric.
Retry with Socratic prompts: Loops on failures up to a configurable cap, using targeted Socratic questions to repair missing causal logic.
Artifacts for auditability: Writes an audit trail to .cheese/hard/<slug>.md with statuses (PASS/FAIL/FAILED/LOGGED) and attempts.

Quick Start

Run hard-cheese on your diff using /hard-cheese <slug> to gate sharing until you can explain the causal logic clearly.

Frequently Asked Questions about hard-cheese

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I enforce a causal reasoning check before a code review?▼

You can enforce a causal reasoning check before code review by using a metacognitive gate that requires the author to explain how and why a diff produces its behavior, grading the explanation against a rubric before sharing.

What is epistemic debt in code review and how do I reduce it?▼

Epistemic debt in code review is the accumulated risk from unverified understanding of changes. You reduce it by requiring authors to demonstrate causal understanding of a diff before sharing it, ensuring the logic is articulated and graded.

How do I set up a Socratic prompt retry loop for code review gating?▼

You can set up a Socratic prompt retry loop by configuring a retry cap that triggers targeted Socratic questions on explanation failures, prompting the author to repair missing causal logic before re-evaluation.

Does the fresh-context judge grading use a specific rubric for pass/fail decisions?▼

Yes, the fresh-context judge grading uses a SOLO-based rubric to evaluate explanations. A separate judge sub-agent assesses the author's causal logic and returns a JSON object to determine pass or fail decisions.

Can I log code review gating attempts without blocking the share-for-review boundary?▼

Yes, you can operate in an optional log-only mode at the share-for-review boundary. This records the author's explanation attempts, statuses, and grades to an audit trail artifact without hard-blocking the review.

How do I persist an audit trail for code review gating attempts?▼

You persist an audit trail by writing status records, such as PASS, FAIL, and FAILED, along with attempt counts, to a markdown artifact file, ensuring all explanation evaluations and retry loops are documented for auditability.