Velocity Tracker

Velocity Tracker Skill

You are the velocity-tracker skill. When invoked, you analyze historical project data to provide quantitative metrics for PM decision-making.

When to Invoke This Skill

Invoke this skill when:

• After completing each task group (to track cycle time)
• Before spawning new developers (to check velocity capacity)
• When a task appears stuck (to detect 99% rule violations)
• Before sending BAZINGA (to record final metrics for learning)
• Planning next iteration (to use historical data)

Do NOT invoke when:

• First run with no completed tasks yet
• Emergency bug fixes (skip metrics to save time)
• User explicitly requests fast mode
• No PM state file exists yet

Your Task

When invoked:

1. Execute the velocity tracking script
2. Read the generated metrics report
3. Return a summary to the calling agent

Step 1: Execute Velocity Tracking Script

Use the Bash tool to run the pre-built tracking script.

On Unix/macOS:

bash .claude/skills/velocity-tracker/scripts/track.sh

On Windows (PowerShell):

pwsh .claude/skills/velocity-tracker/scripts/track.ps1

Cross-platform detection: Check if running on Windows ($env:OS contains "Windows" or uname doesn't exist) and run the appropriate script.

This script will:

• Read PM state from database (bazinga/bazinga.db)
• Calculate current velocity and cycle times
• Compare against historical metrics
• Detect trends (improving/stable/declining)
• Identify 99% rule violations
• Generate bazinga/artifacts/{SESSION_ID}/skills/project_metrics.json
• Update bazinga/artifacts/{SESSION_ID}/skills/historical_metrics.json

Step 2: Read Generated Report

Use the Read tool to read the generated report:

bazinga/artifacts/{SESSION_ID}/skills/project_metrics.json

Extract key information:

• current_run.velocity - Story points completed
• current_run.percent_complete - Progress percentage
• trends.velocity - Velocity trend
• warnings - Any 99% rule violations
• recommendations - Action items

Step 3: Return Summary

Return a concise summary to the calling agent:

Velocity Metrics:
- Current velocity: {velocity} story points (historical avg: {historical_avg})
- Trend: {improving/stable/declining}
- Cycle time: {avg_minutes} minutes average
- Completion: {percent}%

{If warnings exist:}
⚠️  Warnings:
- {warning 1}
- {warning 2}

Top Recommendations:
1. {recommendation 1}
2. {recommendation 2}
3. {recommendation 3}

Details saved to: bazinga/artifacts/{SESSION_ID}/skills/project_metrics.json

Example Invocation

Scenario 1: Normal Progress Check

Input: PM agent requests velocity check after completing task group G002

Expected output:

Velocity Metrics:
- Current velocity: 12 story points (historical avg: 10.5)
- Trend: improving
- Cycle time: 45 minutes average
- Completion: 40%

Top Recommendations:
1. Current velocity exceeds historical average - good progress
2. Estimated remaining time: 3.5 hours

Details saved to: bazinga/artifacts/{SESSION_ID}/skills/project_metrics.json

Scenario 2: 99% Rule Violation Detected

Input: PM agent checks why task G003 is taking too long

Expected output:

Velocity Metrics:
- Current velocity: 8 story points (historical avg: 10.5)
- Trend: declining
- Cycle time: 125 minutes average
- Completion: 30%

⚠️  Warnings:
- 99% Rule Violation: Task G003 taking 3x longer than expected (135 min vs 45 min estimate)

Top Recommendations:
1. Escalate G003 to Tech Lead for investigation
2. Velocity below average - may need task breakdown
3. Database tasks pattern: taking 2.5x estimate

Details saved to: bazinga/artifacts/{SESSION_ID}/skills/project_metrics.json

Error Handling

If script fails:

• Check the script output for error messages
• Verify database bazinga/bazinga.db exists and contains PM state
• Return error message to calling agent

If no completed groups:

• Return: "No completed groups yet. Run velocity tracker after completing at least one task group."

If historical data missing:

• Script will create initial baseline
• Note: "First run - no historical comparison available"

Notes

• The script handles all calculation logic (260+ lines)
• Supports both bash (Linux/Mac) and PowerShell (Windows)
• Gracefully degrades if historical data is missing
• Automatically updates historical metrics for future runs