feat: add open-code-review-delegate skill - #408
Conversation
Add a delegation-mode skill where OCR provides deterministic engineering (file selection, rule resolution) and the host agent drives the actual review. This addresses the large-changeset coverage gap without replacing or undermining the existing code-review-and-quality skill. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
I'll be direct rather than string this along. The redesign moves the integration point but not the question already settled in #327: whether this pack carries a first-class skill for one vendor's product. Delegate mode still means the catalog documents your CLI (npm install -g @alibaba-group/open-code-review as a prerequisite, two subcommands, eight shared flags) and inherits every change to that surface. Addy was explicit in #327 that he keeps the pack's skill files clear of specific third-party products, because one pointer invites the next, and a full catalog skill is more of that, not less. So from my side this is a pass on the same grounds as #327, and I don't think another restructuring changes it. The right home for this is your own repo's plugin and skill distribution, which already exists. Catalog additions are Addy's ( @addyosmani ) call, so I'll leave the formal close to him. |
|
Closing this one. It's a skill built around delegating to a specific external tool (Open Code Review), which is the vendor-specific direction I keep the core pack clear of, the same line as the other tool-integration skills. The lighter "See Also" pointer in #327 is about as far as I want a third-party review tool referenced from inside a skill, and even that is still open rather than merged; a standalone delegate skill is a step past it. Genuinely appreciate the work, it would just live better as your own skill than in the core repo. |
Summary
skills/open-code-review-delegate/SKILL.mdas an independent skillcode-review-and-qualityfor large changesets (10+ files)Motivation
Addresses the coverage gap identified in #327 discussion with a different approach:
browser-testing-with-devtools— external tool provides a capability the agent otherwise lacks (reliable file-level coverage for large changesets)Key differences from #327
delegate previewanddelegate ruleTest plan
🤖 Generated with Claude Code