goodvibes-codebase-review（Goodvibes 代码库审查）

Goodvibes Codebase Review & Remediation

Systematic codebase analysis with parallelized remediation using goodvibes agents.

CRITICAL: Be objective & completely honest. No sugar coating, EVER.

• Honest Review: Honest reviews yield better results. Never consider emotions or feelings in your assessments.

Filesystem Boundaries

CRITICAL: Write-local, read-global.

• WRITE/EDIT/CREATE: ONLY within the current working directory and its subdirectories. This is the project root. All changes must be git-trackable.
• READ: Can read any file anywhere for context (node_modules, global configs, other projects for reference, etc.)
• NEVER WRITE to: parent directories, home directory, system files, other projects, anything outside project root.

The working directory when you were spawned IS the project root. Stay within it for all modifications.

MCP Tool Checklist (MANDATORY)

STOP. Before doing ANYTHING, complete this checklist.

CRITICAL PREREQUISITE: You MUST call mcp-cli info <server>/<tool> BEFORE ANY mcp-cli call <server>/<tool>. This is a BLOCKING REQUIREMENT.

Task Start

# ALWAYS check schema first
mcp-cli info plugin_goodvibes_analysis-engine/detect_stack
mcp-cli info plugin_goodvibes_registry-engine/recommend_skills
mcp-cli info plugin_goodvibes_project-engine/project_issues

# Then make the calls
mcp-cli call plugin_goodvibes_analysis-engine/detect_stack '{}'
mcp-cli call plugin_goodvibes_registry-engine/recommend_skills '{"task":"codebase review quality audit"}'
mcp-cli call plugin_goodvibes_project-engine/project_issues '{}'

Review Phase

# Check schemas first
mcp-cli info plugin_goodvibes_analysis-engine/find_circular_deps
mcp-cli info plugin_goodvibes_analysis-engine/scan_for_secrets
mcp-cli info plugin_goodvibes_project-engine/analyze_dependencies
mcp-cli info plugin_goodvibes_analysis-engine/find_dead_code
mcp-cli info plugin_goodvibes_project-engine/get_test_coverage

# Then execute
mcp-cli call plugin_goodvibes_analysis-engine/find_circular_deps '{}'
mcp-cli call plugin_goodvibes_analysis-engine/scan_for_secrets '{}'
mcp-cli call plugin_goodvibes_project-engine/analyze_dependencies '{}'
mcp-cli call plugin_goodvibes_analysis-engine/find_dead_code '{}'
mcp-cli call plugin_goodvibes_project-engine/get_test_coverage '{}'

Before Every Edit

# Check schemas
mcp-cli info plugin_goodvibes_analysis-engine/scan_patterns
mcp-cli info plugin_goodvibes_project-engine/find_tests_for_file
mcp-cli info plugin_goodvibes_analysis-engine/validate_edits_preview

# Execute
mcp-cli call plugin_goodvibes_analysis-engine/scan_patterns '{}'
mcp-cli call plugin_goodvibes_project-engine/find_tests_for_file '{"file":"path/to/file.ts"}'
mcp-cli call plugin_goodvibes_analysis-engine/validate_edits_preview '{}'

After Every Edit

# Check schema
mcp-cli info plugin_goodvibes_project-engine/project_issues

# Execute
mcp-cli call plugin_goodvibes_project-engine/project_issues '{}'

THE LAW: If a goodvibes MCP tool can do it, USE THE TOOL. No exceptions.

Precision Tools (MANDATORY)

CRITICAL: Use precision tools, NOT system tools.

Precision Tool Patterns

# Find files with pattern (minimal output)
precision_grep:
  queries:
    - pattern: "TODO|FIXME|HACK"
      glob: "src/**/*.ts"
  output:
    mode: files_only

# Read file structure without content
precision_read:
  files: ["src/index.ts", "src/app.ts"]
  extract: outline
  output:
    mode: minimal

# Batch edit multiple files atomically
precision_edit:
  edits:
    - file: "src/api.ts"
      find: "const API_URL = 'http://localhost'"
      replace: "const API_URL = process.env.API_URL"
    - file: "src/config.ts"
      find: "debug: true"
      replace: "debug: process.env.NODE_ENV !== 'production'"
  transaction:
    mode: atomic
    rollback_on_fail: true
  output:
    mode: minimal

# Execute commands with expectations
precision_exec:
  commands:
    - cmd: "npm run typecheck"
      expect:
        exit_code: 0
    - cmd: "npm run lint"
      expect:
        exit_code: 0
  output:
    mode: minimal

Discovery -> Batch Workflow

CRITICAL: Always discover before batching.

The discover tool runs multiple queries in parallel to gather context before building a batch. This prevents wasted operations and ensures you target exactly the right files.

Discovery Tool Usage

# Run parallel discovery queries
discover:
  queries:
    - id: find_components
      type: glob
      patterns: ["src/components/**/*.tsx"]
    - id: find_api_routes
      type: glob
      patterns: ["src/api/**/*.ts", "src/app/api/**/*.ts"]
    - id: find_issues
      type: grep
      pattern: "TODO|FIXME|HACK"
      glob: "src/**/*.{ts,tsx}"
    - id: find_hooks
      type: symbols
      query: "use"
      kinds: ["function"]
  output_mode: files_only  # count_only | files_only | locations

Workflow Pattern

1. Discover - Run queries to understand scope

- Use count_only first to gauge magnitude - Then files_only to get target list

1. Plan - Build batch operations using discovery results

- Reference discovered files in batch operations - Scope work to exactly what was found

1. Execute - Run batch with full context

Example: Feature Implementation

# Step 1: Discover current state
discover:
  queries:
    - id: existing_files
      type: glob
      patterns: ["src/features/auth/**/*.ts"]
    - id: existing_patterns
      type: grep
      pattern: "export (function|const|class)"
      glob: "src/features/**/*.ts"
  output_mode: files_only

# Step 2: Use results to build targeted batch
batch:
  id: implement-feature
  operations:
    read:
      - id: analyze
        type: files
        targets: "{{existing_files.files}}"  # From discovery
        extract: outline

Benefits:

• Prevents blind operations on wrong files
• Ensures consistent patterns across the codebase
• Reduces token usage by targeting exactly what's needed
• Enables informed decisions about implementation approach

Mode-Aware Behavior

Your behavior adapts based on the current mode:

vibecoding Mode

• Communicate: Show progress, explain decisions, report results in detail
• Ask: On ambiguity or risk, ask the user before proceeding
• Checkpoint: Create checkpoints per batch
• Output: Standard verbosity, show diffs

justvibes Mode

• Silent: Minimal communication, log to .goodvibes/logs/activity.md
• Autonomous: Make best-guess decisions, proceed with checkpoints on risk
• Auto-chain: Continue to next logical batch automatically
• Output: Minimal verbosity, no diffs

Workflow Overview

┌─────────────────────────────────────────────────────────────┐
│                      CODEBASE REVIEW                        │
├─────────────────────────────────────────────────────────────┤
│  Phase 1: Review      │  Run MCP tools, analyze all code   │
│  Phase 2: Report      │  Generate codebase-review-report.md │
│  Phase 3: Plan        │  Create remediation-plan.md        │
│  Phase 4: Execute     │  WORK-REVIEW-FIX-CHECK (max 6)     │
└─────────────────────────────────────────────────────────────┘

Phase 1: Codebase Review

MCP Tool Mapping

Review Checklist

Minimum per category:

• 5 specific file:line findings OR explicit "no issues found"
• Quantified measurements (counts, percentages, LOC)
• Severity classification (critical/high/medium/low)

Phase 2: Master Report

Generate codebase-review-report.md:

# Codebase Review Report

**Project**: {name from detect_stack}
**Stack**: {technologies detected}
**Generated**: {ISO 8601 timestamp}
**Overall Score**: {X.X}/10

## Executive Summary

| Severity | Count | Description |
|----------|-------|-------------|
| 🔴 Critical | N | Issues requiring immediate attention |
| 🟠 High | N | Issues blocking production readiness |
| 🟡 Medium | N | Issues impacting maintainability |
| 🔵 Low | N | Minor improvements |

## Score Breakdown

| Category | Weight | Raw | Deductions | Score | Grade |
|----------|--------|-----|------------|-------|-------|
| Quality | 15% | 10 | -X.X | X.X | A-F |
| Architecture | 15% | 10 | -X.X | X.X | A-F |
| Security | 20% | 10 | -X.X | X.X | A-F |
| Performance | 10% | 10 | -X.X | X.X | A-F |
| Documentation | 5% | 10 | -X.X | X.X | A-F |
| Testing | 15% | 10 | -X.X | X.X | A-F |
| Config | 5% | 10 | -X.X | X.X | A-F |
| Dependencies | 5% | 10 | -X.X | X.X | A-F |
| Errors | 5% | 10 | -X.X | X.X | A-F |
| Style | 5% | 10 | -X.X | X.X | A-F |
| **TOTAL** | 100% | - | - | **X.X** | **X** |

## Score Calculation

{Show the math for each category deduction}

## Detailed Findings

### {Category Name}

#### Finding: {Title}

| Field | Value |
|-------|-------|
| **Severity** | critical\|high\|medium\|low |
| **Location** | `file:line` or `file:startLine-endLine` |
| **Measurement** | {exact number or percentage} |
| **Threshold** | {acceptable value} |
| **Impact** | {business/technical consequence} |
| **Points Deducted** | {-X.X from Category} |

**Evidence:**

{code snippet or tool output}

**Required Fix:**

{specific remediation code or steps}

---

{repeat for all findings}

## What's Working Well

{List genuinely good patterns found, with file references}

## Improvement Roadmap

| Phase | Focus | Expected Impact | New Score |
|-------|-------|-----------------|-----------|
| 1 | Critical fixes | +X.X points | X.X |
| 2 | High priority | +X.X points | X.X |
| 3 | Medium priority | +X.X points | X.X |
| 4 | Polish | +X.X points | X.X |

Phase 3: Remediation Plan

Generate remediation-plan.md:

# Remediation Plan

**Generated**: {timestamp}
**Total Tasks**: {N}
**Execution Strategy**: WORK-REVIEW-FIX-CHECK with parallel goodvibes agents (max 6 concurrent)

## Execution Rules

| Rule | Value |
|------|-------|
| Max concurrent agents | 6 |
| Agent type | goodvibes background |
| Context model | Fresh per task (no accumulation) |
| Tool priority | MCP tools > bash |
| Monitoring | Self-report via SubagentStop hook |
| Workflow | WORK → REVIEW → PASS/FAIL → (FIX → CHECK) |

## Task Definitions

### Wave 1: Critical [P0] - Execute Immediately

#### TASK-001: {Description}

| Field | Value |
|-------|-------|
| **Severity** | critical |
| **Target Files** | `file1.ts`, `file2.ts` |
| **Related Finding** | {reference to report finding} |
| **MCP Tools** | `tool1`, `tool2` |
| **Estimated Duration** | Xm |

**Context from Report:**

{relevant finding details}

---

{repeat for all tasks, grouped by wave}

### Wave 2: High [P1]

### Wave 3: Medium [P2]

### Wave 4: Low [P3]

## Dependency Graph

{Show task dependencies if any exist}

Phase 4: Parallel Agent Execution

WORK-REVIEW-FIX-CHECK Workflow

For each remediation task:

  WORK Agent ─────────────────────> REVIEW Agent
  (goodvibes:engineer)              (goodvibes:reviewer)
       │                                  │
       │                                  ├─> PASS ─> Commit ─> Update Log ─> Next Task
       │                                  │
       │                                  └─> FAIL ─> FIX Agent ─> CHECK Agent
                                                         │              │
                                                         │              ├─> PASS ─> Commit
                                                         │              │
                                                         └──────────────┴─> FAIL (loop)

Agent Role Mapping

Waiting for Agents (CRITICAL)

NEVER use these to check agent status:

• tail command on transcript files
• TaskOutput tool
• Any form of polling

INSTEAD, the orchestrator:

1. Spawns agent with run_in_background: true
2. Stops taking actions (turn ends)
3. Receives SubagentStop hook notification when agent completes
4. Hook message appears in context with status

Concurrency Rules

WORK Agent Prompt Template

# Remediation Task: {TASK_ID}

## MCP Tool Checklist (MANDATORY)

**CRITICAL PREREQUISITE:** You MUST call `mcp-cli info <server>/<tool>` BEFORE ANY `mcp-cli call <server>/<tool>`.

Before ANY edit:

Check schemas first

mcp-cli info plugin_goodvibes_analysis-engine/scan_patterns mcp-cli info plugin_goodvibes_project-engine/find_tests_for_file

Then execute

mcp-cli call plugin_goodvibes_analysis-engine/scan_patterns '{}' mcp-cli call plugin_goodvibes_project-engine/find_tests_for_file '{"file":"{TARGET_FILE}"}'

After EVERY edit:

Check schema first

mcp-cli info plugin_goodvibes_project-engine/project_issues

Then execute

mcp-cli call plugin_goodvibes_project-engine/project_issues '{}'

## Assignment

| Field | Value |
| --- | --- |
| Task ID | {TASK_ID} |
| Severity | {SEVERITY} |
| Description | {DESCRIPTION} |
| Files | {FILE_LIST} |

## Context from Report

{FINDING_DETAILS}

## Instructions

1. Complete ONLY this assigned task
2. Use goodvibes MCP tools BEFORE bash
3. Use precision tools (precision_edit, precision_exec) for all operations
4. Run validation tools after every edit
5. If edit causes new errors, fix them before completing

## Tool Priority (MANDATORY)

1. First: Check `mcp-cli info` then `mcp-cli call plugin_goodvibes_*`
2. Second: Use precision tools (precision_edit, precision_read, etc.)
3. Only then: Fall back to standard tools if no MCP/precision tool exists

## Completion Report

When done, output:

{ "task_id": "{TASK_ID}", "status": "success|failed", "files_modified": ["file1.ts", "file2.ts"], "tests_passed": true|false, "type_check_passed": true|false, "notes": "Any relevant context" }

REVIEW Agent Prompt Template

# Review Task: {TASK_ID}

## MCP Tool Checklist (MANDATORY)

**CRITICAL PREREQUISITE:** You MUST call `mcp-cli info <server>/<tool>` BEFORE ANY `mcp-cli call <server>/<tool>`.

Check schemas first

mcp-cli info plugin_goodvibes_project-engine/project_issues mcp-cli info plugin_goodvibes_analysis-engine/scan_patterns mcp-cli info plugin_goodvibes_project-engine/get_test_coverage

Then execute

mcp-cli call plugin_goodvibes_project-engine/project_issues '{}' mcp-cli call plugin_goodvibes_analysis-engine/scan_patterns '{}' mcp-cli call plugin_goodvibes_project-engine/get_test_coverage '{}'

## Assignment

| Field | Value |
| --- | --- |
| Task ID | {TASK_ID} |
| Work Agent | {WORK_AGENT_ID} |
| Files Modified | {FILE_LIST} |

## Context from WORK Agent

{WORK_COMPLETION_REPORT}

## Review Criteria

| Criterion | Weight | Description |
| --- | --- | --- |
| Correctness | 30% | Does it solve the problem correctly? |
| Completeness | 25% | Are all aspects addressed? |
| Quality | 20% | Is the code clean, maintainable? |
| Safety | 15% | No new errors, security issues? |
| Tests | 10% | Tests pass, coverage maintained? |

## Review Process

1. Read all modified files
2. Run MCP validation tools
3. Check for:
  - Type errors
  - Test failures
  - Security issues
  - Code quality issues
  - Incomplete fixes
4. Calculate score based on criteria weights
5. Provide verdict

## Review Output

{ "task_id": "{TASK_ID}", "verdict": "PASS|FAIL", "score": 8.5, "issues": [ { "severity": "critical|major|minor", "file": "path/to/file.ts", "line": 47, "issue": "Description of the issue", "fix_guidance": "Specific guidance on how to fix" } ], "recommendation": "Details for FIX agent if FAIL" }

## Verdict Rules

- **PASS**: Score >= 8.0 AND no critical issues AND all criteria pass
- **FAIL**: Score < 8.0 OR any critical issue OR any criterion fails

**CRITICAL: Be honest. Better to catch issues now than in production.**

FIX Agent Prompt Template

# Fix Task: {TASK_ID} (Iteration {N})

## MCP Tool Checklist (MANDATORY)

**CRITICAL PREREQUISITE:** You MUST call `mcp-cli info <server>/<tool>` BEFORE ANY `mcp-cli call <server>/<tool>`.

Before ANY edit:

Check schemas first

mcp-cli info plugin_goodvibes_analysis-engine/scan_patterns mcp-cli info plugin_goodvibes_project-engine/find_tests_for_file

Then execute

mcp-cli call plugin_goodvibes_analysis-engine/scan_patterns '{}' mcp-cli call plugin_goodvibes_project-engine/find_tests_for_file '{"file":"{TARGET_FILE}"}'

After EVERY edit:

Check schema first

mcp-cli info plugin_goodvibes_project-engine/project_issues

Then execute

mcp-cli call plugin_goodvibes_project-engine/project_issues '{}'

## Review Feedback

{ISSUES_FROM_REVIEW}

## Instructions

1. Address EVERY issue from the review - no skipping
2. Follow the `fix_guidance` provided for each issue
3. Use precision tools for all operations
4. Run validation after each fix
5. If you cannot fix an issue, document WHY and mark as blocked

## Completion Report

{ "task_id": "{TASK_ID}", "iteration": {N}, "status": "success|blocked", "issues_addressed": N, "issues_remaining": N, "blocked_issues": [ { "issue": "Description", "reason": "Why it cannot be fixed" } ], "notes": "Any relevant context" }

CHECK Agent Prompt Template

# Check Task: {TASK_ID} (Verification Round {N})

## MCP Tool Checklist (MANDATORY)

**CRITICAL PREREQUISITE:** You MUST call `mcp-cli info <server>/<tool>` BEFORE ANY `mcp-cli call <server>/<tool>`.

Check schemas first

mcp-cli info plugin_goodvibes_project-engine/project_issues mcp-cli info plugin_goodvibes_analysis-engine/scan_patterns

Then execute

mcp-cli call plugin_goodvibes_project-engine/project_issues '{}' mcp-cli call plugin_goodvibes_analysis-engine/scan_patterns '{}'

## Assignment

| Field | Value |
| --- | --- |
| Task ID | {TASK_ID} |
| FIX Agent | {FIX_AGENT_ID} |
| Iteration | {N} |

## Previous Issues

{ISSUES_TO_VERIFY}

## Context from FIX Agent

{FIX_COMPLETION_REPORT}

## Verification Process

1. Read all modified files
2. Run MCP validation tools
3. Verify each previous issue is resolved
4. Check for new issues introduced by fixes
5. Provide verdict

## Verdict Output

{ "task_id": "{TASK_ID}", "iteration": {N}, "verdict": "PASS|FAIL", "resolved_issues": N, "unresolved_issues": [ { "issue": "Description", "status": "still_present|new_issue" } ], "recommendation": "Next action if FAIL" }

## Verdict Rules

- **PASS**: ALL previous issues resolved AND no new critical issues
- **FAIL**: ANY issue unresolved OR new critical issues found

**If FAIL after 3 iterations, escalate to human review.**

Commit Protocol

After REVIEW or CHECK verdict is PASS, create commit:

fix({CATEGORY}): {TASK_ID} - {SHORT_DESCRIPTION}

{DETAILED_DESCRIPTION}

## Changes

- {FILE_1}: {WHAT_CHANGED}
- {FILE_2}: {WHAT_CHANGED}

## Review

- Reviewed by: REVIEW agent
- Score: {SCORE}/10
- Iterations: {FIX_ITERATIONS}

Task-ID: {TASK_ID} Severity: {SEVERITY} Co-Authored-By: Claude Opus 4.5 [noreply@anthropic.com](https://github.com/mgd34msu/goodvibes-plugin/blob/HEAD/plugins/goodvibes/skills/goodvibes-codebase-review/mailto:noreply@anthropic.com)

Completion Cleanup

After ALL remediation tasks have passed review and been committed, archive the output files:

Archive Location: .goodvibes/completed/

Archive Commands:

# Run all commands together in a single shell session:
# Create archive directory if it doesn't exist
mkdir -p .goodvibes/completed

# Generate timestamp
TIMESTAMP=$(date +%Y%m%d-%H%M%S)

# Move and rename files
mv codebase-review-report.md .goodvibes/completed/code-review-${TIMESTAMP}.md
mv remediation-plan.md .goodvibes/completed/remediation-plan-${TIMESTAMP}.md
mv remediation-log.md .goodvibes/completed/remediation-log-${TIMESTAMP}.md

Archive Checklist:

• All remediation tasks show status: completed in remediation-log.md
• All commits pushed (if applicable)
• Memory files updated (.goodvibes/memory/)
• Archive directory created
• Files moved with timestamp suffix

File Naming Convention:

Why Archive:

• Keeps project root clean
• Preserves historical review data
• Enables tracking of codebase health over time
• Prevents confusion with future reviews

Completion Logging

Maintain remediation-log.md:

# Remediation Log

## Active Agents

| Task ID | Description | Agent | Phase | Started | Elapsed |
|---------|-------------|-------|-------|---------|---------|
| TASK-002 | Remove secrets | agent-abc | WORK | 10:05:00 | 5m |
| TASK-003 | Fix auth flow | agent-def | REVIEW | 10:08:00 | 2m |

## Completed Tasks

| Task ID | Description | Status | Duration | Iterations | Changes |
|---------|-------------|--------|----------|------------|---------|
| TASK-001 | Fix SQL injection | ✅ PASS | 15m32s | 1 | `api/users.ts` |

## Failed Tasks

| Task ID | Description | Status | Iterations | Reason | Retry? |
|---------|-------------|--------|------------|--------|--------|
| TASK-005 | Fix type errors | ❌ FAIL | 3 | Blocked by external dependency | No |

## Summary

- **Total Tasks**: N
- **Completed**: N (X%)
- **In Progress**: N
- **Failed**: N
- **Remaining**: N
- **Active Agents**: N/6

Output Artifacts

Note: After all remediation tasks are completed, these files are archived to .goodvibes/completed/ with timestamp suffixes. See "Completion Cleanup" section for details.

Memory Integration

After review/remediation, update .goodvibes/memory/:

Constraints