Skip to content

feat(skill): code review skill spectrum #103

Description

@KemingHe

Related

Problem

Code review feedback spans multiple levels of abstraction. The skill catalog currently has no structured approach to code review at any level. Issues #101 and #102 address levels 1 and 2, but the broader spectrum needs a roadmap to avoid ad-hoc skill creation and ensure each skill has a clear, non-overlapping scope.

Proposed Spectrum

Level Skill Name Scope Status
1 - Language idioms idiomatic-review Single expression or statement: language-native constructs, built-ins, standard library usage Active (#101)
2 - Code design design-pattern-review Function, class, or component: DI, composition vs inheritance, guard clauses, strategy pattern, error handling patterns Active (#102)
3 - Module design TBD Package boundaries, dependency direction, circular imports, public API surface, layering Parked
4 - Application architecture TBD Whole-app structural decisions: MVC, hexagonal, clean architecture Parked
5 - System design TBD Multi-service, infrastructure, distributed systems Parked

Design Principles

  • Each level is a standalone skill with no dependency on other levels
  • All skills are language-agnostic with auto language detection
  • Composable with other skills (e.g., senior-mentor) via loose coupling
  • SOLID principles cut across levels 2-4 and are referenced within relevant skills rather than forming their own skill
  • Each skill includes a self-contained triage report template for structured output (see cross-cutting concern below)

Cross-Cutting Concern: Triage Report Template

All code review skills in this spectrum should include a structured triage report as their output format:

| Tier | Location | Current Pattern | Suggested Pattern | Reference |
| :--- | :--- | :--- | :--- | :--- |
| High/Medium/Low | File + line range or function/class name | What the code does now | The suggested alternative | Doc link (when available) |

Each skill defines its own triage template independently for self-containment, but implementers should be aware that sibling skills use a parallel format to enable consistent user experience across review levels.

Parked Items

Level 3 - Module design

  • Naming: "module-design-review" is ambiguous across language ecosystems (Python module vs TypeScript module vs Go package)
  • Scope limitation: Module-level review requires whole-codebase context (import graphs, package structure, dependency direction) which is harder for a lightweight skill to deliver reliably without project-specific configuration
  • Action: Revisit naming and feasibility after levels 1-2 are implemented and field-tested

Level 4 - Application architecture

  • Scope limitation: Whole-app structural decisions are highly context-dependent and opinionated, making a generic skill harder to deliver without project-specific configuration
  • Action: Revisit after levels 1-3 are explored

Level 5 - System design

  • Scope limitation: Multi-service and infrastructure concerns extend beyond single-codebase review into deployment, networking, and operational domains
  • Action: Revisit if demand emerges from levels 1-4 usage

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions