Prompt file imported from szabeszcsi/ai-windsurf-workflow (.windsurf/workflows/abort.md). Copyright stays with the author.

name: abort description: Emergency abort and rollback. Usage: /abort auto_execution_mode: 1

Abort Workflow

Emergency recovery when things go wrong.

⚠️ BEFORE EXECUTING THIS: Tell the user what this will do, and ask for confirmation! Also tell the user what are the requirements for this workflow in order to successfully execute the steps!

🚨 STOP IMMEDIATELY

🚨 ABORT INITIATED - STOPPING ALL WORK

Do not make any more changes until recovery path is chosen.

When to Use

• 🔥 Broke something and can't fix it
• 🔥 Tests failing after multiple attempts
• 🔥 Wrong branch, wrong files modified
• 🔥 Context severely degraded
• 🔥 Need to undo recent changes

Step 1: Assess Damage

git status
git diff --stat

Ask user:

• What went wrong?
• How far back do we need to go?
• Are there uncommitted changes we want to keep?

Step 2: Choose Recovery Path

Option A: Discard Uncommitted Changes

git checkout -- .
git clean -fd

⚠️ Destroys all uncommitted work

Option B: Stash Changes for Later

git stash push -m "Recovery stash {date}"

Saves current state, can retrieve later with git stash pop

Option C: Restore from Backup (.bak files)

# If .bak files exist
ls *.bak
mv {file}.bak {file}

Restores from backup files if they were created

Option D: Reset to Last Commit

git reset --hard HEAD

⚠️ Destroys all changes since last commit

Option E: Reset to Specific Commit

git log --oneline -10
git reset --hard {commit-hash}

⚠️ Destroys all changes since that commit

Option F: Selective Restore

# Restore specific file
git checkout HEAD -- path/to/file

# Restore specific directory
git checkout HEAD -- path/to/directory/

Option G: Create Recovery Branch

git checkout -b recovery/{date}

Preserves current mess for later analysis, then switch back to main branch.

Step 3: Verify Recovery

git status
git diff --stat

# Run tests if applicable
pytest tests/ -v --tb=short

Step 4: Document Incident (Optional but Recommended)

Create docs/working/incident_{date}.md:

# Incident {date}

## What Went Wrong
{description of the problem}

## Symptoms
- {symptom 1}
- {symptom 2}

## Recovery Action
{Option X: description}

## Root Cause
{if known}

## Prevention
{what to do differently next time}

Step 5: Report

🔄 ABORT COMPLETE

Recovery: {action taken}
Status: {clean/needs more work}
Branch: {current branch}

Recommendation: Start fresh chat with /start-session

Quick Recovery Commands

After Recovery

1. Take a breath - mistakes happen
2. Start fresh chat - degraded context likely contributed
3. Review what went wrong - learn from it
4. Consider smaller steps - commit more frequently