📊 Copilot Token Usage Report2026-04-14

[github-actions]

Overview

Period: 2026-04-13T23:31Z to 2026-04-14T22:56Z (~23 hours)
Runs analyzed: 42 total (39 with token data; 2 cancelled without data; 1 in-progress)
Total tokens: 12.7M | Total effective tokens: 15.0M
Estimated total cost: N/A (api-proxy estimated_cost not populated)
Total Actions minutes: 220 min
Total turns: 39 across all runs
Total errors: 18 (in 18 runs across 7 workflows)

i️ Cost is not available — estimated_cost is null in all runs. Token counts are from the AWF api-proxy firewall logs. All models: claude-sonnet-4.6.

---

Workflow Summary

Workflow	Runs	Total Tokens	Avg Tokens/Run	Total Eff Tokens	Cache Eff	Errors
Build Test Suite	7	3,103K	443K	4,027K	90%	4/7 (57%) ⚠️
Firewall Issue Dispatcher	4	2,259K	565K	2,519K	93%	1/4 (25%)
Smoke Services	9	1,924K	214K	2,121K	86%	5/9 (56%) ⚠️
Daily Security Review	1	1,830K	1,830K	2,040K	93%	0/1
Smoke Copilot	9	1,205K	134K	1,333K	80%	3/9 (33%) ⚠️
Daily Copilot Token Usage Analyzer	2	693K	347K	789K	88%	0/2
CI/CD Gap Assessment	1	656K	656K	733K	83%	0/1
Dependency Security Monitor	2	437K	218K	483K	76%	0/2
Smoke Chroot	6	383K	64K	558K	71%	4/6 (67%) ⚠️
Issue Duplication Detector	1	177K	177K	434K	90%	1/1 (100%) ❌
Totals	42	12,667K	302K	15,038K	87%	18/42 (43%)

---

🔍 Optimization Opportunities

1. Daily Security Review and Threat Modeling — 1.8M tokens/run (10× threshold)

   • Single run consuming 1.8M tokens is the highest per-run cost in the dataset
   • Consider breaking this into smaller, scoped reviews (e.g., diff-only analysis vs. full codebase scan)
   • High cache efficiency (93%) shows the system prompt is large but stable — consider chunking into multiple scoped passes

2. Build Test Suite — 443K avg tokens/run + 57% error rate ⚠️

   • 443K tokens/run exceeds 100K threshold by 4×
   • 4 out of 7 runs had errors (57% > 30% threshold) — reliability issues may be wasting tokens on failed runs
   • Investigate what's causing failures; failures that consume large token counts before aborting are doubly costly

3. Firewall Issue Dispatcher — 565K avg tokens/run + 1 failure consumed 1.1M tokens ⚠️

   • One failure run (§24385235795) used 1.1M tokens — nearly double the successful runs (243K–603K)
   • High failure token cost suggests the agent processes a lot before hitting an error; add early validation gates

4. CI/CD Gap Assessment — 656K tokens for a single run

   • One-off run consuming 656K tokens; check if this is expected scope or can be narrowed
   • Consider paginating analysis across multiple lower-cost runs if recurring

5. Smoke Chroot / Smoke Services / Smoke Copilot — High error rates (67%, 56%, 33%) ⚠️

   • Smoke tests are failing frequently; failed runs still consume tokens
   • Prioritize fixing underlying test infrastructure issues to reduce wasted token spend
   • Smoke Chroot has lowest cache efficiency (71%) — system prompt may be varying between runs

6. Issue Duplication Detector — 100% error rate (1/1 runs)

   • 177K tokens consumed on a run that failed completely — investigate root cause
   • Very high effective token inflation (177K raw → 434K effective, 2.45× multiplier) suggests unusual cache miss pattern

---

Per-Workflow Run Details

Build Test Suite

• Runs: 7 (§24423923329 cancelled, §24424029722 ✅, §24423591376 cancelled, §24423751335 ✅, §24422261480 ✅, §24421517439 ✅, §24411708818 ✅)
• Total tokens: 3,103K (avg 443K/run across 5 token-reporting runs)
• Total effective tokens: 4,027K (avg 576K)
• Cache efficiency: 90%
• Error rate: 4/7 runs (cancelled runs counted as errors)

Firewall Issue Dispatcher

• Runs: 4 (§24416841845 ✅ 243K, §24399308258 ✅ 603K, §24385235795 ❌ 1,109K, §24375177210 ✅ 304K)
• Total tokens: 2,259K (avg 565K/run)
• Total effective tokens: 2,519K (avg 630K)
• Cache efficiency: 93%
• Error rate: 1/4 runs (25%) — the failed run used 1.1M tokens

Smoke Services

• Runs: 9 (§24424029960 ✅ 304K, §24423923366 cancelled, §24423751362 ✅ 201K, §24423591395 cancelled ✅ 201K, §24422261504 ✅ 238K, §24421517452 ✅ 202K, §24411708813 ✅ 236K, §24399858387 ❌ 271K, §24375289961 ❌ 271K)
• Total tokens: 1,924K (avg 214K/run)
• Total effective tokens: 2,121K
• Cache efficiency: 86%
• Error rate: 5/9 runs (56%) — includes 2 genuine failures + 3 cancellations

Daily Security Review and Threat Modeling

• Runs: 1 (§24400536430 ✅)
• Total tokens: 1,830K | Effective tokens: 2,040K
• Cache efficiency: 93% — good re-use of stable system context
• Error rate: 0/1

Smoke Copilot

• Runs: 9 (§24424029728 ✅, §24423923400 cancelled, §24423751371 ✅, §24423591370 cancelled, §24422261475 ✅, §24421517417 ✅, §24411708834 ✅, §24399187160 ✅, §24375084223 ❌)
• Total tokens: 1,205K (avg 134K/run)
• Cache efficiency: 80%
• Error rate: 3/9 (33%) — 2 cancellations + 1 genuine failure

Daily Copilot Token Usage Analyzer

• Runs: 2 (§24394474136 ✅ 693K, §24427069770 in-progress — current run)
• Total tokens: 693K (1 reporting run)
• Cache efficiency: 88%

CI/CD Pipelines and Integration Tests Gap Assessment

• Runs: 1 (§24399905594 ✅)
• Total tokens: 656K | Effective tokens: 733K
• Cache efficiency: 83%
• One-time analysis workflow — high token count expected for deep codebase assessment

Dependency Security Monitor

• Runs: 2 (§24419923938 ✅ 219K, §24372326443 ✅ 218K)
• Total tokens: 437K (avg 218K/run) — very consistent
• Cache efficiency: 76% — lowest among stable workflows; potential for prompt caching improvement
• Error rate: 0/2

Smoke Chroot

• Runs: 6 (§24424029764 ✅, §24423923313 cancelled, §24423751317 ✅, §24423591372 cancelled, §24422261503 ✅, §24421517581 ✅)
• Total tokens: 383K (avg 64K/run on token-reporting runs, ~96K each)
• Cache efficiency: 71% — lowest in dataset
• Error rate: 4/6 (67%) — primarily cancellations

Issue Duplication Detector

• Runs: 1 (§24385534612 ❌)
• Total tokens: 177K | Effective tokens: 434K (2.45× multiplier — unusual)
• Cache efficiency: 90%
• Error rate: 1/1 (100%) — workflow failed entirely

Workflows Without Token Data

The following runs reported token_usage as null or 0 — likely cancelled before agent reached LLM calls:

Run	Workflow	Status
§24423923329	Build Test Suite	cancelled
§24423591376	Build Test Suite	cancelled
§24427069770	Daily Copilot Token Usage Analyzer	in_progress (current)
§24423923313	Smoke Chroot	cancelled
§24423923366	Smoke Services	cancelled

These are all cancellations — no action required. Cancelled runs that do have token data (e.g., Smoke Copilot cancelled runs) are included in the totals.

---

Historical Trend

Compared to previous report #1971 (2026-04-13T10:44Z → 2026-04-14T10:39Z):

Metric	Previous (#1971)	Current	Change
Total runs	20	42	+110% 🔼
Avg tokens/run	~440K	~302K	-31% ✅
Total tokens	8.8M	12.7M	+44% 🔼
Total effective tokens	10.4M	15.0M	+44% 🔼
Action minutes	191 min	220 min	+15%
Error count	6	18	+200% ⚠️

Key observations:

• Run volume doubled (20 → 42), primarily due to multiple Build Test Suite and Smoke workflow retries/cancellations
• Average tokens per run decreased 31% — efficiency is improving per-run
• Total token consumption rose 44% driven entirely by volume increase (especially multiple smoke test retries)
• Error count tripled to 18 — the spike in errors is a signal worth investigating in CI infrastructure

Previous Report

#1971 — 📊 Copilot Token Usage Report 2026-04-14 (morning run)

---

References:

• §24427069770 — Current workflow run
• §24394474136 — Previous Daily Token Analyzer run
• §24400536430 — Daily Security Review (highest single-run cost)

Generated by Daily Copilot Token Usage Analyzer · ● 521.3K · ◷