review-pr-deeply

Reviews pull requests by delegating parallel multi-perspective analysis to subagents.

Updated Dec 25, 2021
One-click install
npx skills add https://github.com/kotahashihama/dotfiles --skill review-pr-deeply-kotahashihama
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: review-pr-deeply
Source: https://github.com/kotahashihama/dotfiles/tree/main/home/.claude/skills/review-pr-deeply
Command: npx skills add https://github.com/kotahashihama/dotfiles --skill review-pr-deeply-kotahashihama

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Standard PR reviews can miss issues because a single reviewer pass covers limited angles. This Skill replaces only the review-execution step of /review-pr with parallel subagent reviews, giving multi-perspective coverage of a pull request's diff while keeping the rest of the established review workflow intact. ## Core Features & Use Cases - Parallel Review Delegation: Replaces Step 4 of /review-pr by delegating the diff range, convention file paths, and focus points to /review-code-deeply or a project-specific deep review skill. - PR-Specific Cross-Checks: While subagents run, it verifies PR description versus diff consistency, alignment with the PR's stated purpose, and whether previously flagged issues remain. - Severity Remapping: Translates the delegate's severity scale (e.g., must/imo) into the repository's own conventions, requiring evidence from written rules and counted codebase precedents before raising severity. - Use Case: When a teammate asks for a thorough review of a large PR, invoke this Skill to fan out the diff across parallel review agents, then feed the deduplicated findings back into /review-pr's verification and posting steps. ## Quick Start Ask the assistant to deeply review pull request number 123 with a focus on error handling using /review-pr-deeply.

Frequently Asked Questions about review-pr-deeply

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

FAQPage Schema
How do I run a deep multi-perspective review on a GitHub pull request?

Invoke /review-pr-deeply with a PR number or URL and optional focus points. It follows /review-pr for setup, then delegates the diff to parallel review subagents and merges their findings back into the standard verification and posting steps.

What is the difference between /review-pr and /review-pr-deeply?

/review-pr-deeply replaces only Step 4 (review execution) of /review-pr with parallel subagent reviews. All other steps, including PR identification, convention loading, verification, and posting, remain handled by /review-pr.

When should I not use deep parallel PR review?

Skip it for casual requests like 'take a quick look' since it consumes significant compute across multiple agents. Use plain /review-pr for lightweight reviews, and confirm cost with the user when triggered from natural language.

Does this skill work if the project has its own deep review skill?

Yes, and the project-level skill takes priority. It checks .claude/skills/ and .claude/commands/ for a multi-perspective review skill and delegates there first, since project skills know the directory structure and conventions better.

Why do subagent review findings get rejected during verification?

Subagent findings lack verified grounding, so they fail adversarial checks at a higher rate than self-written findings. Each finding must be confirmed against written conventions and counted codebase precedents before reporting.