pr-review-subagents-available

Category: Tools & Productivity | Uploader: inkeep | Downloads: 0 | Version: v1.0（Latest）
Catalog of pr-review-* subagents with their scope, core questions, and model allocation. Use when dispatching PR reviews, improving reviewers, or understanding the review ecosystem.
Changelog: Source: GitHub https://github.com/inkeep/team-skills
Directory Structure

Current level: ci/closed-pr-review-auto-improver/skills/pr-review-subagents-available/
📄 SKILL.md 7.7 KB
SKILL.md

---
name: pr-review-subagents-available
description: |
  Catalog of pr-review-* subagents with their scope, core questions, and model allocation.
  Use when dispatching PR reviews, improving reviewers, or understanding the review ecosystem.
user-invocable: false
disable-model-invocation: true
---

# PR Review Subagents Catalog

## Intent

This skill documents all available `pr-review-*` subagents, their ownership areas, and when to use each one.

**Use this skill when:**
- Dispatching PR reviews (orchestrator needs to know which reviewers to spawn)
- Improving reviewers (need to know which agent owns a pattern)
- Understanding scope boundaries (avoid overlap between agents)

---

## Subagent Roster

### Priority Ranking

Combined ranking by **reasoning demand** (how much model capability matters) and **criticality** (cost of a miss).

| Priority | Agent | Core Question | Model | Reasoning | Criticality |
|----------|-------|---------------|-------|-----------|-------------|
| #1 | `pr-review-security-iam` | "Could this be exploited?" | Opus | Tier 1 | Tier 1 |
| #2 | `pr-review-architecture` | "Will this age well?" | Opus | Tier 1 | Tier 1 |
| #3 | `pr-review-product` | "Is this good for customers?" | Opus | Tier 1 | Tier 2 |
| #4 | `pr-review-breaking-changes` | "Will this break existing users?" | Opus | Tier 3 | Tier 1 |
| #5 | `pr-review-consistency` | "Does this fit the existing world?" | Opus | Tier 2 | Tier 2 |
| #6 | `pr-review-errors` | "Does error handling follow best practices?" | Sonnet | Tier 2 | Tier 3 |
| #7 | `pr-review-standards` | "Is this code correct, secure, and clean?" | Sonnet | Tier 3 | Tier 2 |
| #8 | `pr-review-types` | "Does this type allow illegal states?" | Sonnet | Tier 3 | Tier 3 |
| #9 | `pr-review-tests` | "Are critical paths tested?" | Sonnet | Tier 3 | Tier 3 |
| #10 | `pr-review-frontend` | "Does this follow React/Next.js patterns?" | Sonnet | Tier 3 | Tier 3 |
| #11 | `pr-review-comments` | "Are comments accurate and useful?" | Sonnet | Tier 3 | Tier 3 |
| #12 | `pr-review-docs` | "Is documentation complete and correct?" | Sonnet | Tier 3 | Tier 3 |
| #13 | `pr-review-3p-specs` | "Does this honor the external contract it claims to conform to?" | Opus | Tier 2 | Tier 1 |
| #14 | `pr-review-precision` | "Is every change earning its place — root cause, not symptom?" | Opus | Tier 1 | Tier 2 |
| #15 | `pr-review-llm` | "Is the LLM integration correct, safe, and robust?" | Opus | Tier 2 | Tier 2 |
| #16 | `pr-review-sre` | "Will this survive production traffic and failures?" | Opus | Tier 2 | Tier 2 |
| #17 | `pr-review-devops` | "Will this build, ship, and run correctly?" | Opus | Tier 2 | Tier 3 |

### Tier Definitions

**Reasoning Demand:**
| Tier | What it means |
|------|---------------|
| Tier 1 | Requires reasoning about things *not in the code* — future implications, customer psychology, attacker mindset |
| Tier 2 | Cross-file/cross-surface reasoning grounded in concrete evidence |
| Tier 3 | Pattern matching against loaded skills/checklists/documented rules |

**Criticality:**
| Tier | What it means |
|------|---------------|
| Tier 1 | Misses cause incidents or irreversible debt (security vulnerabilities, data loss, one-way-door architecture) |
| Tier 2 | Misses cause meaningful customer/developer pain (confusing product surface, convention drift, shipped bugs) |
| Tier 3 | Misses are recoverable in follow-up PRs (stale comments, missing tests, suboptimal types) |

---

## Scope Boundaries (Key Distinctions)

### Types vs Consistency

| Concern | Owner | Example |
|---------|-------|---------|
| Type allows illegal states | `pr-review-types` | Optional fields for mutually exclusive states |
| Type duplicates existing schema | `pr-review-consistency` | Manual type when `z.infer` could be used |
| Unsafe type narrowing | `pr-review-types` | `as` assertion without runtime check |
| Type naming doesn't match peers | `pr-review-consistency` | `UserData` when siblings use `*Info` |

### Standards vs Security-IAM

| Concern | Owner | Example |
|---------|-------|---------|
| Generic SQL injection | `pr-review-standards` | User input in query without parameterization |
| Tenant isolation bypass | `pr-review-security-iam` | Missing org_id filter in multi-tenant query |
| XSS vulnerability | `pr-review-standards` | Rendering user input without sanitization |
| Auth bypass | `pr-review-security-iam` | Missing authentication middleware |

### Precision vs Standards vs Architecture

| Concern | Owner | Example |
|---------|-------|---------|
| Fix targets symptom, not root cause | `pr-review-precision` | Null checks in 8 files instead of fixing the source of null |
| Over-engineered infrastructure for simple task | `pr-review-precision` | New validation framework for a single config option |
| Code has a correctness bug | `pr-review-standards` | Off-by-one error in loop |
| Module boundary is wrong | `pr-review-architecture` | Feature in wrong domain package |

### SRE vs Errors vs DevOps

| Concern | Owner | Example |
|---------|-------|---------|
| Missing retry/backoff on external call | `pr-review-sre` | HTTP call without retry strategy |
| Swallowed error in catch block | `pr-review-errors` | `catch (e) {}` with no logging |
| CI workflow misconfiguration | `pr-review-devops` | Action pinned to mutable tag |
| Observability cardinality explosion | `pr-review-sre` | User ID as span attribute |

### LLM vs Standards vs Security-IAM

| Concern | Owner | Example |
|---------|-------|---------|
| Prompt injection risk | `pr-review-llm` | User input concatenated into system prompt |
| General input validation bug | `pr-review-standards` | Missing URL validation on API input |
| Auth bypass in API endpoint | `pr-review-security-iam` | Missing middleware on route |
| Unbounded agent tool loop | `pr-review-llm` | No max iterations on tool-calling loop |

### 3p-Specs vs Architecture vs Consistency

| Concern | Owner | Example |
|---------|-------|---------|
| Implementation diverges from external protocol spec | `pr-review-3p-specs` | SDK wrapper drops required field from external schema |
| Internal module boundary is wrong | `pr-review-architecture` | Feature in wrong domain package |
| Internal naming doesn't match peers | `pr-review-consistency` | Handler not following sibling naming convention |

### Architecture vs Consistency

| Concern | Owner | Example |
|---------|-------|---------|
| Wrong module boundary | `pr-review-architecture` | Feature in wrong domain package |
| File in wrong directory | `pr-review-consistency` | Handler not following peer file placement |
| Circular dependency | `pr-review-architecture` | Packages importing each other |
| Import pattern differs from peers | `pr-review-consistency` | Using default import when peers use named |

---

## Quick Reference: Which Agent for Which Pattern?

| Pattern | Agent |
|---------|-------|
| DRY violations (types, schemas, utilities) | `pr-review-consistency` |
| Type safety / illegal states | `pr-review-types` |
| Auth/authz/tenancy | `pr-review-security-iam` |
| Module boundaries / dependencies | `pr-review-architecture` |
| Error handling patterns | `pr-review-errors` |
| Customer-facing surface changes | `pr-review-product` |
| Breaking changes / migrations | `pr-review-breaking-changes` |
| Code correctness / bugs | `pr-review-standards` |
| Test coverage | `pr-review-tests` |
| React/Next.js patterns | `pr-review-frontend` |
| Documentation quality | `pr-review-docs` |
| Comment accuracy | `pr-review-comments` |
| External spec/protocol fidelity | `pr-review-3p-specs` |
| Root cause vs symptom patching | `pr-review-precision` |
| LLM prompts, tools, streaming, agent loops | `pr-review-llm` |
| Production reliability, retries, observability | `pr-review-sre` |
| CI/CD, dependencies, build/ship pipeline, AI artifacts | `pr-review-devops` |
Comments 0
Latest | Hot
Please login before commenting.
No comments yet. Be the first one!
pr-review-subagents-available

Directory Structure

SKILL.md

Report

Notice
