Overview
Period: 2026-06-27 to 2026-06-28
Runs analyzed: 13 (13/13 had AIC data)
Total AIC: 710.8 units across all workflows
Total Actions minutes: 71 min
Total GitHub API calls: 69
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 (budget cap: 1,000 AIC/day per awf-config.json).
Workflow Summary
| Workflow |
Runs |
Total AIC |
Avg AIC |
Actions Min |
GH API Calls |
Errors |
| Red-Team Benchmark |
1 |
66.3 |
66.3 |
11 |
10 |
0 |
| Smoke Claude |
12 |
644.6 |
53.7 |
60 |
59 |
0 |
All 13 runs completed with success conclusion. No failures recorded.
🔍 Optimization Opportunities
-
🆕 Red-Team Benchmark appeared for the first time — 66.3 AIC, 10.4 min duration
- This is the highest-AIC single run today, exceeding the Smoke Claude average (~53.7)
- Only 1 run in the window (schedule-triggered), so no avg trend yet — monitor over upcoming days
- Recommendation: Establish a baseline over 3–5 runs before flagging as high-cost
-
Bimodal AIC pattern persists in Smoke Claude — 3 runs averaged ~37.2 AIC vs. 9 runs at ~59.2 AIC
- All 3 low-AIC runs are
pull_request-triggered (same finding as yesterday)
- Two bursts of concurrent PR runs at nearly the same second: 15:49 UTC (×2) and 16:17 UTC (×2)
- The concurrent pairs likely test the same or overlapping code paths but yield very different AIC — one path exits earlier
- Recommendation: Add
concurrency: { group: "smoke-${{ github.head_ref }}", cancel-in-progress: true } to avoid redundant same-branch runs
-
Total AIC +8.4% vs. yesterday (655.6 → 710.8)
- The Smoke Claude per-run average is slightly lower today (53.7 vs. 54.6 yesterday)
- The total increase is entirely due to Red-Team Benchmark being added — no organic inflation
- No action required on Smoke Claude; monitor Red-Team Benchmark AIC in subsequent runs
Per-Run Details
Smoke Claude — 12 Runs
Red-Team Benchmark — 1 Run
| Time (UTC) |
AIC |
Event |
Actions Min |
GH API Calls |
Run |
| 2026-06-27 17:16 |
66.3 |
schedule |
11 |
10 |
§28296222524 |
Workflows Without Token Data
All 13 runs have token_usage as null/0. The aic metric is a normalized cost proxy from the AWF api-proxy. To enable per-model token breakdowns (input/output/cache tokens and USD cost estimates), the runs would need to emit structured token usage through the api-proxy's token tracking capability.
Affected workflows: Smoke Claude, Red-Team Benchmark
Historical Trend
| Date |
Runs |
Total AIC |
Avg AIC/Run |
Actions Min |
| 2026-06-23 |
0 |
0 |
— |
0 |
| 2026-06-24 |
18 |
446.8* |
— |
102 |
| 2026-06-25 |
6 |
329.5 |
54.9 |
35 |
| 2026-06-26 |
16 |
749.0 |
46.8 |
85 |
| 2026-06-27 |
12 |
655.6 |
54.6 |
57 |
| 2026-06-28 |
13 |
710.8 |
54.7 |
71 |
* 2026-06-24 had 9 uninstrumented runs; AIC reflects only instrumented subset.
The Smoke Claude avg AIC has been stable at ~54–55 over the last 4 days. The step up in total AIC today is fully explained by Red-Team Benchmark's debut.
Previous Report
📊 Claude Token Usage Report2026-06-27 (#5596)
References:
Generated by Daily Claude Token Usage Analyzer · 68.3 AIC · ⊞ 8.6K · ◷
Overview
Period: 2026-06-27 to 2026-06-28
Runs analyzed: 13 (13/13 had AIC data)
Total AIC: 710.8 units across all workflows
Total Actions minutes: 71 min
Total GitHub API calls: 69
Workflow Summary
All 13 runs completed with
successconclusion. No failures recorded.🔍 Optimization Opportunities
🆕 Red-Team Benchmark appeared for the first time — 66.3 AIC, 10.4 min duration
Bimodal AIC pattern persists in Smoke Claude — 3 runs averaged ~37.2 AIC vs. 9 runs at ~59.2 AIC
pull_request-triggered (same finding as yesterday)concurrency: { group: "smoke-${{ github.head_ref }}", cancel-in-progress: true }to avoid redundant same-branch runsTotal AIC +8.4% vs. yesterday (655.6 → 710.8)
Per-Run Details
Smoke Claude — 12 Runs
Red-Team Benchmark — 1 Run
Workflows Without Token Data
All 13 runs have
token_usageas null/0. Theaicmetric is a normalized cost proxy from the AWF api-proxy. To enable per-model token breakdowns (input/output/cache tokens and USD cost estimates), the runs would need to emit structured token usage through the api-proxy's token tracking capability.Affected workflows: Smoke Claude, Red-Team Benchmark
Historical Trend
* 2026-06-24 had 9 uninstrumented runs; AIC reflects only instrumented subset.
The Smoke Claude avg AIC has been stable at ~54–55 over the last 4 days. The step up in total AIC today is fully explained by Red-Team Benchmark's debut.
Previous Report
📊 Claude Token Usage Report2026-06-27 (#5596)
References: