Task Graph Reporting

Generate progress reports, analyze metrics, track costs and velocity across task-graph projects.

Prerequisite: Understand task-graph-basics for tool reference.

Quick Start

# Connect (reporting doesn't need special tags)
connect(name="reporter", tags=["reporting"])
→ agent_id

# Get project overview
list_tasks(format="markdown")

# Analyze specific task tree
get(task=root_task_id, children=true, format="markdown")

# Check agent activity
list_agents(format="markdown")

Report Types

1. Status Report

Quick overview of project state:

┌─────────────────────────────────────────┐
│ STATUS REPORT                           │
├─────────────────────────────────────────┤
│ Query: list_tasks(format="markdown")    │
│                                         │
│ Metrics to extract:                     │
│ • Total tasks by status                 │
│ • Blocked tasks and blockers            │
│ • Active agents and their tasks         │
│ • Ready tasks (available work)          │
└─────────────────────────────────────────┘

2. Progress Report

Track completion over time:

┌─────────────────────────────────────────┐
│ PROGRESS REPORT                         │
├─────────────────────────────────────────┤
│ Queries:                                │
│ • list_tasks(status="completed")        │
│ • list_tasks(status="working")      │
│ • list_tasks(status="pending")          │
│                                         │
│ Calculate:                              │
│ • Completion rate (completed/total)     │
│ • Points completed vs remaining         │
│ • Time actual vs estimated              │
└─────────────────────────────────────────┘

3. Cost Report

Analyze resource consumption:

┌─────────────────────────────────────────┐
│ COST REPORT                             │
├─────────────────────────────────────────┤
│ Query: get(task=root, children=true)    │
│                                         │
│ Aggregate across tasks:                 │
│ • tokens_in, tokens_out, tokens_cached  │
│ • tokens_thinking, tokens_image/audio   │
│ • cost_usd total and per-task           │
└─────────────────────────────────────────┘

4. Velocity Report

Measure team throughput:

┌─────────────────────────────────────────┐
│ VELOCITY REPORT                         │
├─────────────────────────────────────────┤
│ Queries:                                │
│ • list_tasks(status="completed")        │
│ • get_state_history(task=task_id)       │
│                                         │
│ Calculate:                              │
│ • Points completed per time period      │
│ • Average time per point                │
│ • Agent productivity comparison         │
└─────────────────────────────────────────┘

5. Agent Report

Analyze agent activity:

┌─────────────────────────────────────────┐
│ AGENT REPORT                            │
├─────────────────────────────────────────┤
│ Query: list_agents(format="markdown")   │
│                                         │
│ Per agent:                              │
│ • Current claims                        │
│ • Tasks completed                       │
│ • Time since last heartbeat             │
│ • Tags (capabilities)                   │
└─────────────────────────────────────────┘

Metrics Reference

Task Metrics

Cost Metrics

Agent Metrics

Query Patterns

By Status

# All completed
list_tasks(status="completed")

# Multiple statuses
list_tasks(status=["pending", "working"])

# Only ready (unclaimed, unblocked)
list_tasks(ready=true)

# Only blocked
list_tasks(blocked=true)

By Hierarchy

# Root tasks only
list_tasks(parent="null")

# Children of specific task
list_tasks(parent=task_id)

# Full tree
get(task=root_id, children=true)

By Owner

# Specific agent's tasks
list_tasks(owner=agent_id)

# Unclaimed only
list_tasks(owner="null", status="pending")

Time-Based (via state history)

# Get state transitions for a task
get_state_history(task=task_id)

# Returns:
# - Each state entered and exited
# - Duration in each state
# - Agent who made transitions

Report Templates

Executive Summary

# Project Status: {project_name}
Generated: {timestamp}

## Overview
- **Total Tasks:** {total}
- **Completed:** {completed} ({percent}%)
- **In Progress:** {working}
- **Blocked:** {blocked}

## Velocity
- **Points Completed:** {points_done} / {points_total}
- **Avg Time per Point:** {avg_time}

## Cost
- **Total Cost:** ${total_cost}
- **Cost per Point:** ${cost_per_point}

## Active Agents
| Agent | Tasks | Last Active |
|-------|-------|-------------|
{agent_rows}

## Blockers
{blocked_tasks_list}

Burndown Data

# Query completed tasks over time
# Plot: remaining points vs time

Day 1: {total_points}
Day 2: {total_points - completed_day_2}
Day 3: {total_points - completed_day_3}
...

Cost Breakdown

# Cost Report: {project_name}

## By Task
| Task | Tokens In | Tokens Out | Cost |
|------|-----------|------------|------|
{task_rows}

## By Agent
| Agent | Tasks Done | Total Cost |
|-------|------------|------------|
{agent_rows}

## Totals
- **Total Tokens:** {sum_tokens}
- **Total Cost:** ${sum_cost}

Analysis Patterns

Bottleneck Detection

1. list_tasks(blocked=true)
2. For each blocked task, identify blocker
3. Group by blocker → find most-blocking tasks
4. Priority = blocked_count × blocked_priority

Estimation Accuracy

1. list_tasks(status="completed")
2. For each: accuracy = time_actual / time_estimate
3. Calculate mean, median, std deviation
4. Flag tasks with accuracy < 0.5 or > 2.0

Agent Utilization

1. list_agents()
2. For each agent:
   - Current claims = active work count
   - Time since last_heartbeat = idle_time
3. Flag: utilization 0 or idle_time > threshold

Generating Reports

Manual Report

# 1. Connect
connect(name="reporter") → agent_id

# 2. Gather data
tasks = list_tasks(format="markdown")
agents = list_agents(format="markdown")

# 3. For cost data, traverse tree
root = get(task=root_id, children=true)

# 4. Aggregate and format
# (Calculate totals, percentages, etc.)

Automated Report (via attachment)

# Store report as attachment on root task
attach(
  task=root_id,
  name="weekly-report",
  content=report_markdown,
  mime="text/markdown"
)

Best Practices

Reporting Frequency

Data Hygiene

• Ensure workers log costs consistently
• Verify time tracking is enabled (timed states)
• Check for orphaned tasks (no parent, not root)
• Validate estimates exist for velocity calcs

Related Skills