void-code-review

Reviews pull request diffs across six dimensions with blocker and nit classification.

Updated May 29, 2026
One-click install
npx skills add https://github.com/voidcorp-core/void-harness --skill void-code-review-voidcorp-core
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: void-code-review
Source: https://github.com/voidcorp-core/void-harness/tree/main/packages/core/skills/void-code-review
Command: npx skills add https://github.com/voidcorp-core/void-harness --skill void-code-review-voidcorp-core

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Code reviews without a framework degrade into superficial comments that miss correctness, security, and architectural defects. This Skill enforces a structured pre-PR review pass so every diff is checked against six ordered dimensions with explicit blocker versus nit classification and recorded evidence. ## Core Features & Use Cases - Six-dimension review framework: Walks correctness, tests, security, structure, readability, and performance in priority order, composing with specialized skills like void-tdd, void-security-guidance, and void-typescript-strict. - Blocker vs nit discipline: Every comment is prefixed BLOCKER, NIT, QUESTION, or PRAISE, and strict mode requires a Review Evidence block in the PR body. - Cross-model second opinion: Composes the native /void-code-review command at configurable effort levels with the codex CLI for an independent review from a different model family, surfacing disagreements explicitly. - Use Case: Before merging a feature branch into main, run the strict-mode pass to verify TDD evidence, trust-boundary validation, and hexagonal boundaries, then record the dimension checklist and blocker count in the PR body. ## Quick Start Review the current branch diff against main in strict mode and produce the Review Evidence block for the PR body.

Frequently Asked Questions about void-code-review

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

FAQPage Schema
How do I run a structured code review on a pull request diff?

Run the review pass on the diff, which walks six dimensions in order: correctness, tests, security, structure, readability, and performance. Each finding is tagged BLOCKER, NIT, QUESTION, or PRAISE, and strict mode produces a Review Evidence block for the PR body.

What is the difference between strict and souple review modes?

Strict mode applies to PRs targeting main or release branches and requires all six dimensions checked, blockers failing the merge, and a Review Evidence block. Souple mode is for in-progress feedback on feature branches with dimensions checked at user discretion and no evidence block required.

How do I get a second opinion from a different AI model on a code review?

The skill composes the codex CLI's non-interactive review subcommand for an independent cross-model pass on high-stakes diffs. Disagreements between Claude and Codex are surfaced explicitly in the evidence block rather than silently arbitrated.

When should a pull request be split before review?

PRs larger than roughly 400 lines of diff should be split into focused PRs, since review quality decays past that threshold. If the change is truly atomic, add a large-cl-justification marker in a commit message instead.

What are the limitations of this code review approach?

The skill orchestrates review dimensions but does not decide whether to ship, does not block on style or naming handled by formatters, and does not perform deep security audits itself. Deep doctrine and security passes are delegated to the doctrine-critic agent and void-security-audit skill.