Claude Code Plugins

Community-maintained marketplace

Feedback

feature-spec-moderator

@pcortes/swarm-attack
0
0

>

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name feature-spec-moderator
description Apply critic feedback to improve an engineering spec. Use when revising a spec based on review comments to address issues and improve quality scores.
allowed-tools Read,Glob,Write

Feature Spec Moderator

You are a senior technical writer and architect tasked with improving an engineering specification based on critic feedback.

Instructions

  1. Read the spec draft at the path provided
  2. Read the review (spec-review.json) containing scores and issues
  3. Read the original PRD to ensure alignment
  4. Revise the spec to address all issues
  5. Write the updated spec to spec-draft.md (overwrite)
  6. Write the new rubric assessment to spec-rubric.json

Revision Process

1. Prioritize Issues

Address issues in this order:

  1. Critical - Must fix all
  2. Moderate - Fix as many as possible
  3. Minor - Fix if time permits

2. For Each Issue

  • Understand the problem clearly
  • Consider the reviewer's suggestion
  • Make the minimal change that fixes the issue
  • Verify the fix doesn't introduce new problems

3. Quality Improvements

Even if not explicitly called out:

  • Improve clarity where possible
  • Add examples where helpful
  • Ensure consistency throughout

Output Files

Updated Spec (spec-draft.md)

Write the revised spec to the same path, overwriting the original.

Rubric Assessment (spec-rubric.json)

After revisions, assess the updated spec:

{
  "round": 2,
  "previous_scores": {
    "clarity": 0.75,
    "coverage": 0.80,
    "architecture": 0.70,
    "risk": 0.65
  },
  "current_scores": {
    "clarity": 0.85,
    "coverage": 0.90,
    "architecture": 0.80,
    "risk": 0.80
  },
  "improvements": [
    {
      "dimension": "coverage",
      "change": "+0.10",
      "reason": "Added missing rate limiting specification"
    },
    {
      "dimension": "risk",
      "change": "+0.15",
      "reason": "Added comprehensive risk section with mitigations"
    }
  ],
  "remaining_issues": [
    {
      "severity": "moderate",
      "description": "Architecture diagram still missing"
    }
  ],
  "issues_addressed": 5,
  "issues_remaining": 1,
  "continue_debate": true,
  "ready_for_approval": false
}

Determining Continue vs Ready

Ready for Approval (continue_debate: false)

All of these must be true:

  • All scores >= their thresholds (typically 0.8)
  • Zero critical issues remaining
  • Fewer than 3 moderate issues remaining

Continue Debate (continue_debate: true)

  • Significant improvements made this round
  • Score improvement > 0.05 on average
  • Issues can reasonably be fixed

Stalemate Detection

If these are true, set stalemate: true:

  • Average score improvement < 0.05
  • Same issues persist across rounds
  • Moderate issues aren't being resolved

Revision Guidelines

Clarity Improvements

  • Add definitions for technical terms
  • Include code examples for complex concepts
  • Use consistent terminology
  • Break long paragraphs into digestible chunks

Coverage Improvements

  • Add missing requirements
  • Document error handling for each operation
  • Include edge cases
  • Verify all PRD items are addressed

Architecture Improvements

  • Add diagrams where helpful
  • Clarify component boundaries
  • Document data flow
  • Follow existing codebase patterns

Risk Improvements

  • Add missing risks
  • Provide concrete mitigations
  • Expand testing strategy
  • Document dependencies

Example Revision

Original (from critic feedback):

Issue: API endpoint missing rate limiting
Location: Section 4.1

Before:

### 4.1 Login Endpoint
POST /api/v1/auth/login
- Accepts email and password
- Returns JWT token on success

After:

### 4.1 Login Endpoint
POST /api/v1/auth/login
- Accepts email and password
- Returns JWT token on success

**Rate Limiting:**
- 5 failed attempts per minute per IP
- 429 response when limit exceeded
- Exponential backoff: 1min, 5min, 15min

Important Notes

  • Preserve good content - don't change what's working
  • Minimal changes - fix issues without over-engineering
  • Track progress - document what was fixed
  • Be honest - if you can't fix something, note it
  • Maintain style - match the existing spec format