📊 Daily Copilot Token Consumption Report - December 23, 2025

[github-actions[bot]]

Executive Summary

Over the last 4 days (December 20-23, 2025), Copilot-powered agentic workflows consumed 159.6 million tokens at an estimated cost of $4.79, across 402 workflow runs covering 46 unique workflows.

Key Highlights:

• Highest consuming workflow: Hourly CI Cleaner (46.9M tokens, $1.41)
• Most active workflow: AI Moderator (99 runs, 3.2M tokens)
• Average cost per run: $0.0119
• Peak day: December 22 with 85.5M tokens across 204 runs
• Trend: Activity stabilizing after weekend spike

Full Report Details

📈 Token Usage Trends

Daily Breakdown (Last 4 Days)

Date	Total Tokens	Total Cost	Runs	Avg Tokens/Run
2025-12-20	654,077	$0.0196	2	327,039
2025-12-21	56,056,288	$1.6817	139	403,284
2025-12-22	85,475,980	$2.5643	204	419,000
2025-12-23	17,429,196	$0.5229	57	305,774

Trend Analysis: Token consumption spiked significantly on December 21-22 (weekdays) with 139-204 runs per day, representing the typical workday load. December 23 shows reduced activity as we approach the holidays. Weekend baseline (Dec 20) shows minimal activity with only 2 runs.

Activity Patterns

• Weekday Average: ~56-85M tokens/day across 139-204 runs
• Weekend Baseline: ~650K tokens/day across 2 runs
• Peak Run Count: December 22 with 204 workflow executions
• Most Efficient Day: December 23 with lowest avg tokens/run (305,774)

🏆 Top 15 Workflows by Token Consumption

Rank	Workflow	Total Tokens	Total Cost	Runs	Avg Tokens/Run	Avg Cost/Run
1	Hourly CI Cleaner	46,927,294	$1.4078	48	977,652	$0.0293
2	Issue Monster	13,511,787	$0.4054	45	300,262	$0.0090
3	Tidy	12,888,995	$0.3867	41	314,366	$0.0094
4	Spec-Kit Execute	10,176,973	$0.3053	10	1,017,697	$0.0305
5	Smoke Copilot No Firewall	7,570,618	$0.2271	19	398,454	$0.0120
6	Dev Hawk	7,111,856	$0.2134	21	338,660	$0.0102
7	Sub-Issue Closer	5,270,787	$0.1581	4	1,317,697	$0.0395
8	Agent Performance Analyzer - Meta-Orchestrator	4,392,137	$0.1318	4	1,098,034	$0.0329
9	The Daily Repository Chronicle	4,147,827	$0.1244	1	4,147,827	$0.1244
10	Q	3,349,417	$0.1005	3	1,116,472	$0.0335
11	AI Moderator	3,219,807	$0.0966	99	32,523	$0.0010
12	Layout Specification Maintainer	3,172,015	$0.0952	2	1,586,008	$0.0476
13	Smoke Copilot	2,677,551	$0.0803	13	205,965	$0.0062
14	Documentation Noob Tester	2,673,293	$0.0802	3	891,098	$0.0267
15	jsweep - JavaScript Unbloater	2,348,583	$0.0705	2	1,174,292	$0.0352

Top 5 Analysis

1. Hourly CI Cleaner (29.4% of total cost)

• Consumes 977K tokens per run (2.5x repository average)
• Runs hourly (48 times over 4 days)
• Single largest consumer at $1.41 (29% of all costs)

2. Issue Monster (8.5% of total cost)

• Moderate token usage: 300K per run
• High frequency: 45 runs (scheduled workflow)
• Second highest total cost at $0.41

3. Tidy (8.1% of total cost)

• Similar profile to Issue Monster
• 41 runs with 314K tokens per run
• Maintenance workflow running regularly

4. Spec-Kit Execute (6.4% of total cost)

• High per-run consumption: 1.0M tokens
• Lower frequency: only 10 runs
• Fourth highest total despite fewer runs

5. Smoke Copilot No Firewall (4.7% of total cost)

• Moderate per-run: 398K tokens
• 19 runs (testing workflow)
• Fifth highest consumer

📊 Cost Distribution

Top 5 workflows account for 57.0% of total costs:

• Hourly CI Cleaner: $1.41 (29.4%)
• Issue Monster: $0.41 (8.5%)
• Tidy: $0.39 (8.1%)
• Spec-Kit Execute: $0.31 (6.4%)
• Smoke Copilot No Firewall: $0.23 (4.7%)

Remaining 41 workflows: $2.06 (43.0%)

💡 Insights & Recommendations

High-Cost Workflow Analysis

1. Hourly CI Cleaner - $1.41 (29% of total)

Observation: This scheduled workflow consumes nearly 1M tokens per run, executing 48 times over 4 days (hourly). Its per-run consumption is 2.5x the repository average.

Recommendations:

• Reduce frequency: Consider running every 2-3 hours instead of hourly during off-peak times
• Optimize prompts: Review if the agent is performing redundant analysis on each run
• Batch operations: Accumulate smaller cleanups and process in larger batches
• Potential savings: Reducing frequency to every 2 hours could save $0.70/4 days ($5/month)

2. Issue Monster + Tidy - $0.80 combined (16.6% of total)

Observation: Both are high-frequency scheduled workflows with moderate per-run consumption (300K-314K tokens). Combined they run 86 times over 4 days.

Recommendations:

• Consolidate operations: Consider if these workflows could share some analysis
• Smart triggering: Only run when there are actual issues/items to process
• Cache recent analysis: Avoid re-analyzing unchanged content
• Potential savings: Smart triggering could reduce runs by 30-40%, saving ~$0.25/4 days

3. High Per-Run Consumers (>1M tokens/run)

Several workflows consume over 1M tokens per run:

• The Daily Repository Chronicle: 4.1M tokens (1 run)
• Sub-Issue Closer: 1.3M tokens (4 runs)
• Agent Performance Analyzer: 1.1M tokens (4 runs)
• Spec-Kit Execute: 1.0M tokens (10 runs)

Recommendations:

• Scope analysis: These workflows may be analyzing too much context
• Incremental processing: Break large operations into smaller chunks
• Context windowing: Limit how much history/context is loaded
• Potential savings: 30% reduction in context could save ~$0.30/4 days

Efficiency Opportunities

1. AI Moderator - High Efficiency Example

Stats: 99 runs, only 32K tokens/run, $0.0010/run

• Most frequent workflow but very token-efficient
• Total cost only $0.10 despite 99 runs
• Good model for other high-frequency workflows

Pattern to replicate:

• Focused scope per execution
• Minimal context loading
• Quick decision-making

2. Scheduled vs. Event-Driven Workflows

Observation: Scheduled workflows (Hourly CI Cleaner, Issue Monster, Tidy) dominate costs because they run regardless of whether work exists.

Recommendation:

• Convert appropriate scheduled workflows to event-driven triggers
• Use GitHub Actions conditional execution based on actual changes
• Implement "no-op" detection early in workflow to exit if nothing to do

Cost Trend Analysis

4-Day Period Breakdown:

• Weekend (Dec 20): $0.02 - Minimal activity
• Monday (Dec 21): $1.68 - 86x increase as workflows resume
• Tuesday (Dec 22): $2.56 - Peak activity, 52% increase
• Monday (Dec 23): $0.52 - 80% decrease (holiday week)

Projection: Typical workweek consumption appears to be ~$1.50-$2.50/day, with weekends at ~$0.02/day.

Monthly estimate: ~$35-50/month during active development periods

🎯 Priority Optimization Recommendations

Immediate Actions (High Impact, Low Effort)

1. Reduce Hourly CI Cleaner frequency

   • Change from hourly to every 2-3 hours during off-peak
   • Estimated savings: $5-7/month
   • Effort: Configuration change only

2. Add no-op detection to scheduled workflows

   • Exit early if no work to do
   • Estimated savings: $3-5/month
   • Effort: Add 5-10 lines of check logic

3. Review high per-run consumers for context optimization

   • Target: Sub-Issue Closer, Agent Performance Analyzer, Daily Chronicle
   • Estimated savings: $2-4/month
   • Effort: Prompt engineering review

Medium-Term Actions (High Impact, Medium Effort)

4. Consolidate similar maintenance workflows

   • Combine Issue Monster + Tidy where logical
   • Estimated savings: $3-4/month
   • Effort: Workflow refactoring

5. Implement smart triggering for scheduled workflows

   • Convert to event-driven where possible
   • Estimated savings: $5-8/month
   • Effort: Rearchitect triggers

Long-Term Monitoring

6. Establish token budgets per workflow

   • Alert when workflows exceed expected consumption
   • Prevents regression after optimization

7. Regular efficiency reviews

   • Monthly review of top 10 consumers
   • Continuous optimization culture

📅 Historical Comparison

Metric	Dec 20	Dec 21	Dec 22	Dec 23	Total
Total Tokens	654K	56.1M	85.5M	17.4M	159.6M
Total Cost	$0.02	$1.68	$2.56	$0.52	$4.79
Total Runs	2	139	204	57	402
Avg Cost/Run	$0.010	$0.012	$0.013	$0.009	$0.012

Key Insights:

• 53.5% of all tokens consumed on December 22 (peak activity day)
• Weekend shows 99% reduction in activity
• Average cost per run is remarkably consistent ($0.009-$0.013)
• Holiday week (Dec 23) shows 80% reduction from peak

🔧 Methodology

• Data Source: GitHub Actions workflow run artifacts from last 30 days (402 runs analyzed)
• Engine Filter: Copilot engine workflows only
• Cost Model: $0.00003 per 1,000 tokens (estimated Copilot pricing)
• Analysis Period: December 20-23, 2025 (4 days)
• Cache Storage: /tmp/gh-aw/cache-memory/token-metrics/history.jsonl
• Data Quality: All 402 runs had complete token usage data

📊 Data Quality Notes

• Complete coverage: 100% of Copilot workflow runs in period had token metrics
• No filtering applied: All workflows included regardless of success/failure
• Cost estimates: Based on rough approximation; actual costs may vary
• Runs included: All workflow executions including re-runs and manual triggers

📦 Repository Statistics

• Total unique workflows: 46 Copilot-powered workflows
• Active workflows: 36 workflows executed at least once in 4 days
• Scheduled workflows: 11 high-frequency scheduled workflows identified
• Event-driven workflows: 35 workflows triggered by repository events
• Most reliable: 41 workflows with >90% success rate

---

Next Steps:

1. Review and implement high-priority optimizations
2. Monitor token consumption after changes
3. Next report: December 26, 2025 (after holiday period)

References:

• Hourly CI Cleaner Example Run §20458754870
• Issue Monster Example Run §20457416267
• AI Moderator Example Run §20458209195

AI generated by Daily Copilot Token Consumption Report

[github-actions[bot]]

This discussion was automatically closed because it was created by an agentic workflow more than 3 days ago.