Agent skill
shared-worker-retrospective
Retrospective contribution format for Developer, Tech Artist, QA, and Game Designer workers. Use proactively when retrospective is triggered to contribute your perspective.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/shared-worker-retrospective
SKILL.md
Worker Retrospective Contributions
"When retrospective triggers, contribute your perspective – read memory, write separate file, delete memory."
When to Use This Skill
Use when:
prd.json.agents.{agent}.status == "awaiting_retrospective".claude/session/retrospective.txtexists
Use proactively:
- MANDATORY: Read ALL task memory files first
- Create your own contribution file (separate file per agent)
- Delete ALL task memory files after contributing
Quick Start
Example 2: QA contribution
// File: .claude/session/retrospective-qa.json
{
"taskId": "feat-001",
"agent": "qa",
"timestamp": "2026-01-23T16:00:00Z",
"contribution": {
"validationResultsSummary": {
"typeScript": "pass",
"lint": "pass",
"tests": "pass",
"build": "pass",
"browser": "pass"
},
"codeQualityObservations": [
"Code is maintainable with clear structure",
"Proper error handling throughout"
],
"qualityConcerns": [],
"suggestionsForImprovement": [
"Consider adding more edge case tests"
]
}
}
Example 3: Tech Artist contribution
// File: .claude/session/retrospective-techartist.json
{
"taskId": "vis-002",
"agent": "techartist",
"timestamp": "2026-01-23T14:30:00Z",
"contribution": {
"visualAssetsCreated": [
"Custom shader for water effects",
"Particle system for explosions"
],
"visualQualityAssessment": "Matches GDD specifications",
"performanceMetrics": {
"frameRateImpact": "minimal",
"drawCalls": "+15",
"shaderComplexity": "medium"
},
"whatWorkedWell": [
"Shader compilation succeeded on first try"
],
"lessonsLearned": [
"Test shaders on mobile targets early"
]
}
}
Contribution Workflow
┌─────────────────────────────────────────────────────────────┐
│ 1. READ ALL TASK MEMORY (MANDATORY) │
│ Directory: .claude/session/agents/{agent}/ │
│ Pattern: task-*.md │
└─────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 2. READ retrospective.txt for context │
└─────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 3. CREATE YOUR CONTRIBUTION FILE │
│ .claude/session/retrospective-{agent}.json │
└─────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 4. DELETE ALL TASK MEMORY FILES │
│ .claude/session/agents/{agent}/task-*.md │
└─────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 5. UPDATE STATUS, LOG, EXIT │
│ PM will merge all contributions │
└─────────────────────────────────────────────────────────────┘
Why Separate Files?
Previous Problem: All workers writing to single retrospective.txt
- Race conditions: last write wins
- File watcher triggers multiple times
- No atomic write guarantees
Solution: Each agent writes to their own file
- No contention between agents
- PM reads all files and merges atomically
- Clear ownership of each contribution
Contribution File Locations
| Agent | Contribution File |
|---|---|
| Developer | .claude/session/retrospective-developer.json |
| Tech Artist | .claude/session/retrospective-techartist.json |
| QA | .claude/session/retrospective-qa.json |
| Game Designer | .claude/session/retrospective-gamedesigner.json |
Contribution Structure
All contributions follow this JSON structure:
{
"taskId": "string",
"agent": "developer|techartist|qa|gamedesigner",
"timestamp": "ISO 8601",
"contribution": {
// Agent-specific fields
}
}
Agent-Specific Fields
Developer
| Field | Description |
|---|---|
implementationDecisions |
Key technical decisions made |
technicalChallenges |
Difficulties faced and solutions |
whatWorkedWell |
Effective solutions/patterns |
areasForImprovement |
What could be better |
lessonsLearned |
Suggestions for future tasks |
QA
| Field | Description |
|---|---|
validationResultsSummary |
TypeScript, lint, test, build, browser results |
codeQualityObservations |
Maintainability, smells, error handling |
qualityConcerns |
Performance, test coverage, patterns |
suggestionsForImprovement |
What would make code better |
Tech Artist
| Field | Description |
|---|---|
visualAssetsCreated |
Materials, shaders, effects implemented |
visualQualityAssessment |
How well visuals match GDD |
performanceMetrics |
Frame rate, draw calls, triangles, memory |
challengesFaced |
Shader/asset integration difficulties |
whatWorkedWell |
Effective visual techniques |
lessonsLearned |
Shader patterns, pipeline improvements |
Game Designer
| Field | Description |
|---|---|
designDecisionsMade |
Key design choices |
gddClarityIssues |
Unclear specifications |
playtestFindings |
Mechanics needing refinement |
whatWorkedWell |
Effective design patterns |
lessonsLearned |
Design process improvements |
Contribution Guidelines
Be Specific
- Mention specific files, functions, patterns
- Note unexpected issues
- Share what surprised you
Be Honest
- If you took shortcuts, mention them
- If something felt hacky, say so
- If PRD was unclear, explain why
Be Constructive
- Suggest improvements for future tasks
- Note what would have helped
- Identify areas needing refactoring
Anti-Patterns
❌ DON'T:
- Skip contributing to retrospective
- Write generic/vague contributions
- Edit other agents' sections
- Delete or modify retrospective structure
- FORGET to read task memory first
- FORGET to delete ALL task memory files
✅ DO:
- Read ALL task memory files first
- Write specific, detailed contributions
- Use your own contribution file
- Delete memory after contributing
- Be honest and constructive
⚠️ Critical Reminder
Your retrospective contribution will be GENERIC and USELESS without reading task memory first!
Task memory files contain everything you noted during execution. Read them ALL before writing your contribution.
Related Skills
| Skill | Purpose |
|---|---|
shared-worker-task-memory |
Task memory tracking during execution |
shared-ralph-core |
Session structure and state management |
shared-ralph-event-protocol |
Event-driven messaging protocol |
Didn't find tool you were looking for?