---

name: story-execution description: "Execute a single user story from prd.json. Use during Ralf loop iterations to implement stories correctly."

Story Execution Skill

Expert guidance for implementing a single user story in one Ralf iteration.

When to Use

• During Ralf loop iterations
• When implementing a story from prd.json
• When you need to ensure proper execution flow

Execution Flow

1. Pre-Implementation

• Read .ralf/prd.json to get story details
• Read .ralf/progress.txt for codebase patterns
• Check you're on the correct branch
• Identify the target story (first with passes: false)

2. Implementation

• Implement ONLY the acceptance criteria listed
• Follow existing code patterns
• Keep changes minimal and focused
• Don't add extras not in acceptance criteria

3. Verification

• Run typecheck (required for all stories)
• Run linter if configured
• Run tests if relevant
• For UI stories: verify visually if possible

4. Post-Implementation

• Commit with message: feat: [US-XXX] - [Story Title]
• Update .ralf/prd.json: set passes: true
• Append to .ralf/progress.txt with learnings
• Update AGENTS.md if reusable patterns found

Quality Gates

A story is only complete when:

• All acceptance criteria met
• Typecheck passes
• No linting errors
• Changes committed
• .ralf/prd.json updated
• .ralf/progress.txt logged

Common Mistakes to Avoid

1. Scope Creep: Implementing more than the acceptance criteria
2. Skipping Verification: Not running typecheck/tests
3. Forgetting Progress: Not updating .ralf/progress.txt
4. Breaking Changes: Changing code outside story scope
5. Missing Commits: Forgetting to commit before loop ends

Progress Entry Format

Copy## [Date] - [Story ID]
- What was implemented
- Files changed
- **Learnings:**
  - Patterns discovered
  - Gotchas encountered
---

Completion Check

After completing a story:

1. Check if ALL stories have passes: true
2. If yes: output <promise>COMPLETE</promise>
3. If no: end response (next iteration continues)