Overview
Period: last 24 hours (2026-09-08 to 2026-09-09)
Runs analyzed: 3 (1 had token data)
Total tokens: 4,910 across all workflows
Estimated total cost: N/A (no estimated_cost field present in this dataset)
Total Actions minutes: 25 min
Only one Claude-engine workflow ran in this period: Smoke Claude. Of its 3 runs, 2 failed with driver_exit and produced no token data; the 1 successful run used 4,910 tokens.
Workflow Summary
| Workflow |
Runs |
Total Tokens |
Avg Tokens |
Cost |
Avg Cost |
Turns |
| Smoke Claude |
3 |
4,910 |
1,637 (per successful run: 4,910) |
N/A |
N/A |
0 (not reported) |
🔍 Optimization Opportunities
- Smoke Claude — reliability, not cost, is the issue. 2 of 3 runs (67%) failed with
failure_kind: "driver_exit" and 0 token usage recorded, exceeding the >30% error-rate threshold. This dwarfs any token-efficiency concern — investigate why the Claude driver process is exiting before completing (check #5808 and #5807 run logs for stack traces / OOM / auth errors).
- No cost or turn data available —
estimated_cost, turns, and token_usage_summary (model/cache breakdown) were absent or zero for all runs. This suggests the api-proxy sidecar's cost/cache instrumentation isn't populating for this workflow, or the successful run was too short to produce a full usage summary. Recommend verifying --enable-api-proxy and Anthropic proxy logging are active on smoke-claude.lock.yml.
- Token volume itself (4,910 tokens on the one successful run) is low and not a cost concern — no action needed there.
Per-Workflow Details
Smoke Claude
- Runs: 3 — §34297918918 (success), §34233952128 (failure/driver_exit), §34228036877 (failure/driver_exit)
- Total tokens: 4,910 (only the successful run reported usage)
- Estimated cost: N/A —
estimated_cost not present in the source data
- Turns: not reported (
turns field absent/0 for all runs)
- Model breakdown: not available —
token_usage_summary was null for all 3 runs
- Cache analysis: not available — no cache read/write data present
- Error rate: 2/3 runs (67%) — both failures classified as
driver_exit with error_count: 2 each
Workflows Without Token Data
- Smoke Claude — 2 of 3 runs (the two
driver_exit failures) had token_usage absent/0, since the agent process exited before usage could be recorded.
Historical Trend
This is the first Claude token usage report found for this repository (no prior claude-token-usage-report-labeled issues located).
Previous Report
None found.
References:
Generated by Daily Claude Token Usage Analyzer · copilot · auto · 23.3 AIC · ⊞ 13.4K · ◷
Overview
Period: last 24 hours (2026-09-08 to 2026-09-09)
Runs analyzed: 3 (1 had token data)
Total tokens: 4,910 across all workflows
Estimated total cost: N/A (no
estimated_costfield present in this dataset)Total Actions minutes: 25 min
Only one Claude-engine workflow ran in this period: Smoke Claude. Of its 3 runs, 2 failed with
driver_exitand produced no token data; the 1 successful run used 4,910 tokens.Workflow Summary
🔍 Optimization Opportunities
failure_kind: "driver_exit"and 0 token usage recorded, exceeding the >30% error-rate threshold. This dwarfs any token-efficiency concern — investigate why the Claude driver process is exiting before completing (check#5808and#5807run logs for stack traces / OOM / auth errors).estimated_cost,turns, andtoken_usage_summary(model/cache breakdown) were absent or zero for all runs. This suggests the api-proxy sidecar's cost/cache instrumentation isn't populating for this workflow, or the successful run was too short to produce a full usage summary. Recommend verifying--enable-api-proxyand Anthropic proxy logging are active onsmoke-claude.lock.yml.Per-Workflow Details
Smoke Claude
estimated_costnot present in the source dataturnsfield absent/0 for all runs)token_usage_summarywas null for all 3 runsdriver_exitwitherror_count: 2eachWorkflows Without Token Data
driver_exitfailures) hadtoken_usageabsent/0, since the agent process exited before usage could be recorded.Historical Trend
This is the first Claude token usage report found for this repository (no prior
claude-token-usage-report-labeled issues located).Previous Report
None found.
References: