📊 Claude Token Usage Report2026-06-27

[github-actions]

Overview

Period: 2026-06-26 to 2026-06-27
Runs analyzed: 12 (12/12 had AIC data)
Total AIC: 655.6 units across all runs
Total Actions minutes: 57 min
Total GitHub API calls: 58

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	12	655.6	54.6	57	58	0

All 12 runs completed with success conclusion. No failures recorded.

🔍 Optimization Opportunities

1. Bimodal AIC distribution — 3 runs averaged ~37 AIC vs. 9 runs at ~60 AIC

   • The lower-AIC runs (~37) are all pull_request-triggered and appear to test a shorter code path or a branch where less work is performed
   • One concurrent pair fired at 19:03 UTC: §28259294932 (AIC=61.6) and §28259326215 (AIC=37.3) — different PRs triggering different task branches
   • Recommendation: Investigate whether the ~37 AIC path can be confirmed as intentional (e.g., a simpler task variant) or if it indicates incomplete agent execution

2. High PR trigger rate — 10 of 12 runs were triggered by pull_request events (vs. 2 by schedule)

   • Previous report had 16 runs with no trigger breakdown available for comparison
   • If PRs are triggering redundant smoke tests on every push, consider restricting to PR open/sync or adding concurrency groups to cancel stale runs
   • Recommendation: Add concurrency: group: smoke-${{ github.head_ref }}, cancel-in-progress: true to the workflow if not already present

3. Slight AIC increase (+2.1% avg) — avg AIC rose from 53.5 → 54.6

   • Within normal variation; trend remains flat across 4 consecutive daily reports
   • No action required unless the trend continues

Per-Run Details

Smoke Claude — All 12 Runs

Time (UTC)	AIC	Event	Actions Min	GH API Calls	Run
2026-06-26 13:13	56.7	schedule	5	10	§28240351157
2026-06-26 15:19	61.4	pull_request	4	3	§28247412047
2026-06-26 15:20	37.2	pull_request	5	3	§28247463098
2026-06-26 15:51	61.4	pull_request	5	4	§28249227373
2026-06-26 16:36	61.4	pull_request	5	4	§28251654120
2026-06-26 16:38	37.2	pull_request	4	4	§28251751541
2026-06-26 18:18	61.4	pull_request	5	4	§28256982627
2026-06-26 19:03	61.7	pull_request	5	4	§28259294932
2026-06-26 19:03	37.3	pull_request	5	4	§28259326215
2026-06-26 19:34	61.6	pull_request	5	4	§28260843234
2026-06-26 21:08	61.5	pull_request	5	4	§28265416951
2026-06-27 01:21	56.7	schedule	4	10	§28274255656

AIC summary: min=37.2, max=61.7, mean=54.6, median=61.4

By trigger type:

• schedule (2 runs): avg AIC = 56.7, avg GH API calls = 10.0
• pull_request (10 runs): avg AIC = 54.2, avg GH API calls = 3.8

Schedule-triggered runs average ~5% higher AIC and ~2.5x more GitHub API calls than PR-triggered runs, consistent with schedule runs performing more complete health checks.

Workflows Without Token Data

All 12 runs in this period had non-zero AIC values. No runs are missing token instrumentation.

However, raw token counts (input_tokens, output_tokens, cache_read_tokens, cache_write_tokens) remain unavailable. This data would require the --artifacts all flag when downloading logs. The api-proxy sidecar is confirmed active (AIC data is being recorded).

Historical Trend

Date	Runs	Instrumented	Avg AIC	Total AIC	Actions Min	Errors
2026-06-27 (today)	12	12	54.6	655.6	57	0
2026-06-26	16	14	53.5	749.0	85	4
2026-06-25	6	6	54.9	329.5	35	0
2026-06-24	18	9	49.6	446.8	102	0

• Avg AIC has been stable in the 49–55 range across 4 reporting days
• Error count returned to 0 (previous report had 4 errors)
• Run count decreased from 16 → 12; lower trigger volume on PRs today

Previous Report

📊 Claude Token Usage Report2026-06-26 — Issue #5555

References:

• §28284692448 — this report's workflow run

Generated by Daily Claude Token Usage Analyzer · 47.7 AIC · ⊞ 8.6K · ◷