Feature Description
Add a structured PR review workflow to Claude Code that automates Copilot triage, convention checks, holistic review, and merge operations through a set of coordinated skills and agents.
5 new skills:
/review-workflow <PR#> — Full end-to-end orchestrator (setup → Copilot triage → automated review → self-check → interview reviewer → decision)
/triage-copilot <PR#> — Fetch and classify Copilot review suggestions (incorporate / ignore / discuss)
/review-pr <PR#> — Run code-critic + doc-sync + pr-reviewer agents in parallel
/draft-pr-comment <PR#> — Draft and post structured change-request comments
/approve-pr <PR#> — Pre-flight safety checks, approve, squash-merge, delete branch
2 new agents:
copilot-triage — Classifies each Copilot suggestion against project conventions
pr-reviewer — Holistic review covering correctness, methodology, architecture, tests, performance, completeness, and data impact
Updates to 3 existing agents (code-critic, doc-sync, test-scaffolder):
- Added tool-usage guidance (no command chaining)
- Fixed dependency group references (
--group lint, --group test)
Settings update:
- Ruff format hook updated to use
--group lint
Use Case / Motivation
The project is receiving a growing volume of PRs. This workflow provides a structured, repeatable process for reviewing them — automating convention checks, Copilot triage, and holistic analysis while keeping the reviewer engaged through follow-up questions and a decision gate before any action is taken.
Proposed Implementation
Implementation is complete. All files live under .claude/agents/ and .claude/skills/. No changes to code/ or tests/.
Alternatives Considered
Manual review checklists — rejected because a structured workflow ensures consistent coverage across all review dimensions without relying on memory.
Feature Description
Add a structured PR review workflow to Claude Code that automates Copilot triage, convention checks, holistic review, and merge operations through a set of coordinated skills and agents.
5 new skills:
/review-workflow <PR#>— Full end-to-end orchestrator (setup → Copilot triage → automated review → self-check → interview reviewer → decision)/triage-copilot <PR#>— Fetch and classify Copilot review suggestions (incorporate / ignore / discuss)/review-pr <PR#>— Run code-critic + doc-sync + pr-reviewer agents in parallel/draft-pr-comment <PR#>— Draft and post structured change-request comments/approve-pr <PR#>— Pre-flight safety checks, approve, squash-merge, delete branch2 new agents:
copilot-triage— Classifies each Copilot suggestion against project conventionspr-reviewer— Holistic review covering correctness, methodology, architecture, tests, performance, completeness, and data impactUpdates to 3 existing agents (
code-critic,doc-sync,test-scaffolder):--group lint,--group test)Settings update:
--group lintUse Case / Motivation
The project is receiving a growing volume of PRs. This workflow provides a structured, repeatable process for reviewing them — automating convention checks, Copilot triage, and holistic analysis while keeping the reviewer engaged through follow-up questions and a decision gate before any action is taken.
Proposed Implementation
Implementation is complete. All files live under
.claude/agents/and.claude/skills/. No changes tocode/ortests/.Alternatives Considered
Manual review checklists — rejected because a structured workflow ensures consistent coverage across all review dimensions without relying on memory.