work-review-request

Dispatches a code reviewer subagent against a diff and drives the fix-and-re-review loop.

3|3|Updated Apr 6, 2026
One-click install
npx skills add https://github.com/philipobenito/elelem --skill work-review-request-philipobenito
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: work-review-request
Source: https://github.com/philipobenito/elelem/tree/main/skills/work-review-request
Command: npx skills add https://github.com/philipobenito/elelem --skill work-review-request-philipobenito

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Code changes often ship without an independent review, or get reviewed against the wrong diff range, producing approvals that describe code nobody actually examined. This Skill establishes the correct review range, dispatches a reviewer subagent, and owns the fix-and-re-review loop until no Critical or Important issue remains. ## Core Features & Use Cases - Review Range Establishment: Determines the correct base commit, includes uncommitted working-tree changes, and surfaces untracked files so the reviewer sees the real change. - Reviewer Dispatch with Model Selection: Fills a structured prompt template and dispatches a reviewer subagent at a model tier matched to the change, escalating only on evidence. - Fix-and-Re-Review Loop: Routes findings through work-review-receive, re-dispatches after fixes within a three-dispatch budget, and returns a strict verdict contract (Approved, Issues outstanding, Nothing to review, Range unknown). - Use Case: Before merging a feature branch or committing a bug fix, invoke this Skill so an independent reviewer examines the full diff, findings get fixed, and the fixes themselves are re-reviewed before approval. ## Quick Start Ask the assistant to run the work-review-request skill to review the current uncommitted changes against the feature base commit before opening a pull request.

Frequently Asked Questions about work-review-request

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

FAQPage Schema
How do I get an AI code review of uncommitted changes before a pull request?

Invoke this skill with the base commit your work started from. It diffs the base against the working tree so uncommitted changes are included, lists untracked files separately, and dispatches a reviewer subagent against that exact range.

How does the automated code review loop handle reviewer findings?

Findings are processed through the work-review-receive skill, which verifies each item, pushes back on incorrect ones, and implements valid fixes in severity order. The reviewer is then re-dispatched against the updated range, up to three dispatches per invocation.

Why should I not use HEAD~1 as the base for a code review diff?

HEAD~1 is only correct for single-commit work. On a multi-task feature it reviews just the last commit while reporting on the whole feature, and it errors on a repository's first commit. Use the recorded base commit or git merge-base instead.

What happens if the code review diff is empty?

An empty diff returns Nothing to review, which is a precondition failure, not an approval. The caller must establish where the work actually is and invoke the skill again, since no review has taken place.

When does the review loop stop and escalate to a human?

The loop stops after three dispatches if Critical or Important issues remain, returning Issues outstanding. Clarification requests and escalations from finding processing also go directly to the user rather than triggering another dispatch.