Structured 4-phase workflow for software development: Research → Create → Generate → Execute. Guides AI through discovery, requirements, task generation, and step-by-step implementation with approval gates.
A structured prompt-based system for guiding AI assistants through complete software development workflows, from research to implementation.
This skill implements a 4-phase development workflow that produces versioned documentation and executes tasks with user approval gates. It's designed for building features, content, and designs in a systematic, auditable way.
Choose the appropriate requirements type:
Apply relevant domain standards from `/standards/domains/` and phase-specific standards.
**Task List Structure:**
- Mark parent task complete with `[x]`
- Commit changes to git
- Push to remote
- Wait for user approval before next parent task
When you need user input:
1. Number all questions (1, 2, 3...)
2. Provide multiple-choice options (A, B, C, D...)
3. Enable concise responses like "1A, 2C, 3B"
Example:
```
1. Authentication approach?
A. JWT tokens
B. Session cookies
C. OAuth only
2. Database choice?
A. PostgreSQL
B. MySQL
C. SQLite
```
Load applicable standards from `/standards/` based on `standards-manifest.yml`:
Include a "Standards Compliance" section in all output documents.
```
User: "I want to add user authentication"
AI:
1. Load research.md prompt
2. Apply global + research phase standards
3. Conduct discovery research
4. Generate /tasks/rsd-authentication-v1.md
5. Ask user: "Ready to proceed to requirements phase?"
```
```
User: "Create a PRD for the authentication feature"
AI:
1. Load create-prd.md prompt
2. Reference /tasks/rsd-authentication-v1.md
3. Apply global + code-architecture standards
4. Generate /tasks/prd-authentication-v1.md
5. Ask user: "Ready to generate tasks?"
```
```
User: "Execute the authentication tasks"
AI:
1. Load execute-tasks.md prompt
2. Reference /tasks/tasks-authentication-v1.md
3. Complete Task 0.0 (create feature branch)
4. Mark [x] Task 0.0
5. **STOP and wait for user approval**
6. Continue to Task 1.1 only after approval
```
1. **One sub-task at a time** - Never batch multiple sub-tasks without approval
2. **No overwrites** - Always version documents (v1, v2, v3...)
3. **Stop after each sub-task** - Wait for explicit user approval
4. **Commit after parent completion** - Git commit/push only when all sub-tasks of a parent are done
5. **Apply standards** - Always check standards-manifest.yml for applicable standards
6. **Use /tasks/ directory** - All generated documents go to `/tasks/`
```
/prompts/
research.md
create-prd.md
create-crd.md
create-drd.md
generate-tasks.md
execute-tasks.md
/standards/
standards-manifest.yml
global/
domains/
phases/
teams/
/tasks/
rsd-*.md
prd-*.md
crd-*.md
drd-*.md
tasks-*.md
```
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/ai-workflow-toolkit/raw