📊 Claude Token Usage Report2026-06-30

[github-actions]

Overview

Period: 2026-06-29 to 2026-06-30
Runs analyzed: 15 (15/15 had AIC data, 0 had token-level data)
Total AIC: 719.7 units across all workflows
Total Actions minutes: 82 min
Total GitHub API calls: 66

i️ This report uses aic (Agent Inference Cost) as the cost metric. USD cost estimates and token-level data are not available in the current data format — token_usage is null for all runs. AIC is a normalized inference cost unit reported by the AWF api-proxy.

Workflow Summary

Workflow	Runs	Total AIC	Avg AIC	Actions Min	GH API Calls	Errors
Smoke Claude	15	719.7	48.0	82	66	0

15/15 runs concluded success. No failures recorded this period.

🔍 Optimization Opportunities

1. Bimodal AIC pattern continues — distribution shifted toward high group — two stable clusters persist

   • 5 PR runs averaged 36.3 AIC (low group — all copilot/export-audit-* and upgrade-gh-aw-pre-release branches)
   • 10 runs averaged 53.8 AIC (high group — all other PRs + 2 schedule runs)
   • Yesterday had 7 low + 7 high; today has 5 low + 10 high
   • The low group is consistent: copilot-generated export-related branches consistently exit earlier (simpler task)
   • Recommendation: Instrument Smoke Claude to emit turn count and tool call count per run — this would clarify whether low-AIC runs exit early due to simpler scope or hit a different code path

2. Schedule runs use 3× more GitHub API calls than PR runs — schedule event averages 10 API calls vs. 3.5 for pull_request events

   • Both schedule runs made exactly 10 GitHub API calls (vs. 3–4 for most PR runs)
   • This suggests the schedule path does additional repo scanning or metric collection
   • Recommendation: Review the scheduled smoke test logic to identify if any GH API calls are redundant or can be cached

3. High-AIC group clusters at two sub-levels (52–54 AIC vs. 58+ AIC)

   • Mid-tier ~52 AIC: copilot/duplicate-code-writeconfigs-setup, feat/model-api-mapping, arc-dind-build-tools-sysroot, runner-doctor-*
   • Top-tier ~58 AIC: test/ssl-key-storage-config-writer-coverage, fix/failing-test-imports, copilot/remove-securewipefile-export
   • Test-coverage and fix branches consistently hit the 58 AIC ceiling — likely more complex agent tasks
   • Recommendation: Review if test-coverage workflows can be scoped more narrowly to reduce agent context load

Historical Trend

Date	Runs	Total AIC	Avg AIC	Actions Min	Error Rate
2026-06-30	15	719.7	48.0	82	0% ✅
2026-06-29	16	767.2	48.0	84	6.3%
2026-06-28	13	710.8	54.7*	~67	n/a

* Estimated from prior reports.

Trend: Total AIC is down -6.2% from yesterday (767.2 → 719.7), driven by 1 fewer run. Per-run efficiency is flat at 48.0 AIC/run — consistent for the second day in a row. Error rate improved from 6.3% to 0% (all 15 runs succeeded). No new workflows appeared; Red-Team Benchmark workflow was not triggered in this window.

Per-Run Details — Smoke Claude (15 Runs)

Time (UTC)	AIC	Event	Actions Min	GH API Calls	Conclusion	Run
06-29 13:53	53.7	schedule	5	10	success	§28377163584 · main
06-29 13:58	36.2	pull_request	5	4	success	§28377526731 · upgrade-gh-aw-pre-release
06-29 14:30	58.2	pull_request	6	3	success	§28379600713 · test/ssl-key-storage-config-writer-coverage
06-29 15:19	58.2	pull_request	5	4	success	§28382840893 · copilot/remove-securewipefile-export
06-29 15:20	36.1	pull_request	7	4	success	§28382906555 · copilot/export-audit-remove-resolveprovidersessionid
06-29 15:21	36.1	pull_request	6	4	success	§28382949477 · copilot/export-audit-fix-satisfies
06-29 15:22	36.1	pull_request	6	4	success	§28383032356 · copilot/remove-unused-export-buildprovidertargetenv
06-29 15:24	36.8	pull_request	5	4	success	§28383125571 · copilot/fix-ajv-import-path
06-29 16:46	58.2	pull_request	6	4	success	§28388248127 · fix/failing-test-imports
06-29 21:30	52.2	pull_request	5	3	success	§28403949345 · runner-doctor-b7-eacces-chroot-home
06-29 23:45	52.3	pull_request	5	3	success	§28410182710 · arc-dind-build-tools-sysroot
06-30 01:22	48.7	schedule	6	10	success	§28413808937 · main
06-30 01:54	52.2	pull_request	5	3	success	§28414958720 · copilot/duplicate-code-writeconfigs-setup
06-30 03:57	52.3	pull_request	5	3	success	§28419204122 · copilot/duplicate-code-writeconfigs-setup
06-30 04:34	52.3	pull_request	5	3	success	§28420538920 · feat/model-api-mapping

Workflows Without Token Data

All 15 runs reported token_usage = null. Token-level data (input/output/cache tokens) is not yet available for the Smoke Claude workflow. AIC is the only cost signal available.

To enable token-level reporting:

• Ensure the api-proxy sidecar (--enable-api-proxy) is active and emitting token metrics
• Verify gh aw logs picks up token_usage_summary from the api-proxy logs path

Previous Report

📊 Claude Token Usage Report 2026-06-29

References:

• §28433891254 — this report's workflow run

Generated by Daily Claude Token Usage Analyzer · 50.8 AIC · ⊞ 8.6K · ◷

[github-actions]

This issue is being closed as outdated. A newer issue has been created: #5744

View newer issue

---

This action was performed automatically by the Daily Claude Token Usage Analyzer workflow.