/gemini-scan

Read-only independent code review using Gemini. Reports issues without making any changes. Senior Google engineer style - no handholding, no false praise.

No arguments? Describe this skill and stop. Do not execute.

Craft Standards: What We're Looking For

Code that a master craftsperson would be proud of.

This scan specifically looks for code that appears AI-generated rather than crafted by a skilled human engineer.

AI Antipatterns Gemini Will Flag

Antipattern	Example	Why It's Wrong
Over-abstraction	Factory used once	Abstraction without justification
Defensive paranoia	`if (x !== null)` when x can't be null	Doesn't understand the code
Reimplementing stdlib	Custom `deepClone` when `structuredClone` exists	Ignorance or arrogance
Comment spam	`// loop through users` above `for (user of users)`	Insulting the reader
Speculative features	Config options with one possible value	Solving imaginary problems
Enterprise patterns	`AbstractUserFactoryBean`	Cargo-culting

Goal: Report code that wouldn't survive a senior engineer's review.

Target

If a path argument is provided, review that file/directory. If no argument, review recently modified files (git diff/log). Multiple paths can be provided to scan a set of components.

Process

Step 0: Load Rubric

Read .claude/rubric/AUTO-DETECT.md for the detection table. Then:

•Always load: .claude/rubric/base.md and .claude/rubric/product-quality.md
•Auto-detect domains: Check target files against the detection table. Load matching domain rubrics (.claude/rubric/web-api.md, .claude/rubric/data-persistence.md, .claude/rubric/cli.md, .claude/rubric/microservice.md).
•Extract Review Criteria: From each loaded rubric, collect the numbered items under ## Review Criteria. Combine into a single criteria list for the Gemini context.

If a rubric file doesn't exist, skip it and continue.

Step 1: Find Code to Review

Find target files:

•If path provided, use that
•Otherwise, find recently modified files using git diff or git log
•Look in: src/, lib/, app/, and project root

If NO code exists, output "no code to review" and stop.

Step 2: Read All Target Code

Read ALL files in scope completely. Do not skim.

Step 3: Call Gemini (MANDATORY)

For each file or logical group:

code

mcp__gemini-reviewer__gemini_review
  code: <paste the source code>
  focus: "general"
  context: "PRODUCTION READINESS review. If you wouldn't deploy this to production today, flag it. Review against: {RUBRIC_CRITERIA}. For each finding cite file:line and severity (CRITICAL/HIGH/MEDIUM/LOW). CRITICAL = blocks production. HIGH = would cause incidents."

Replace {RUBRIC_CRITERIA} with the combined Review Criteria from all loaded rubric files, numbered sequentially.

Note: Test Coverage is handled by the testing phase — do not include it in the rubric criteria.

If tool unavailable, output: GEMINI_ERROR: tool not available

Step 4: Compile Report (DO NOT FIX)

Collect all issues from Gemini and organize by severity.

DO NOT edit any files. Report only.

Output Format

markdown

## Gemini Scan: [target]

### Summary

| Metric | Value |
|--------|-------|
| Files scanned | N |
| Total lines | N |
| Critical issues | N |
| High issues | N |
| Medium issues | N |
| Low issues | N |

### Critical Issues 🔴

Must fix before shipping:

1. **[file:line]** — [description]
   - Problem: [what Gemini found]
   - Impact: [why it matters]
   - Suggested fix: [how to address]

### High Issues 🟠

Should fix:

1. **[file:line]** — [description]
   - Problem: [what Gemini found]
   - Suggested fix: [how to address]

### Medium Issues 🟡

Consider fixing:

1. **[file:line]** — [description]
   - Concern: [what Gemini found]

### Low Issues 💭

Minor improvements:

1. **[file:line]** — [description]

### AI-Generated Antipatterns Detected

Patterns typical of AI-generated code that should be simplified:

- [ ] Over-abstraction (factories/wrappers used once)
- [ ] Features not requested
- [ ] Defensive checks for impossible cases
- [ ] Reimplementing stdlib
- [ ] Copy-paste that should be extracted
- [ ] Over-commenting obvious code
- [ ] Unnecessary config options
- [ ] Over-engineered types

### Files Reviewed

| File | Lines | Issues |
|------|-------|--------|
| path/to/file.ts | 245 | 2 🔴, 1 🟠 |
| ... | ... | ... |

---
GEMINI_RESULT: called - [N] total issues
SCAN_ONLY: no fixes applied

Rules

•READ ONLY - Do not edit any files
•MUST CALL GEMINI - This skill requires the Gemini MCP tool
•COMPLETE - Review all files in scope
•ORGANIZE - Group issues by severity
•ACTIONABLE - Include suggested fixes (but don't apply them)

When to Use

•Pre-commit quality check
•Code review before PR
•Assessing unfamiliar code
•Finding issues without auto-fix
•Getting external perspective on your code

Anti-Patterns (Don't Do)

•Making any edits to files
•Skipping the Gemini call
•Summarizing without specific line numbers
•Hiding or downplaying issues
•Applying fixes (use /gemini-review for that)