[spending-forecast] Daily spending forecast - 2026-08-21

[github-actions]

Daily Spending Forecast — github/gh-aw

Forecast date: 2026-08-21T09:54:01Z
History window: 30 days (per-workflow history_days)
Period basis: month
Run reference: §32469658856

Executive Summary

• Total observed AIC (30-day window, 23 active workflows): 38,808.08
• Weekly forecast total — P10 (10th percentile — optimistic scenario: 9 out of 10 weeks will cost at least this much): 2,945.38 | P50 (50th percentile — median/expected scenario): 7,592.57 | P90 (90th percentile — conservative scenario: only 1 out of 10 weeks is expected to exceed this): 13,960.92
• Monthly forecast total — P10 (10th percentile — optimistic): 21,103.00 | P50 (50th percentile — median): 33,629.21 | P90 (90th percentile — conservative): 49,388.27
• Active workflows sampled: 23 of 49 tracked workflows had runs in the 30-day window; 26 had zero sampled runs (dormant/disabled/no-trigger workflows) and are excluded from the projections above.
• Top spender: Go Logger Enhancement — observed 10,389.21 AIC across 28 runs (success rate 75.0%).

Charts

Native PNG chart rendering (matplotlib/pandas) was unavailable in this run's Python environment (package installation failed due to disk space exhaustion on the runner). Falling back to ASCII visual summaries as specified.

Chart 1 — Spending Trend (Last 30 Days, top 5 workflows by total AIC)

Spending Trend — 7-day rolling avg AIC/run (Top 5 Workflows)

Go Logger Enhancement        ▇▇▇▇▇█▇▆▅▅  latest=289.3 avg=371.0
Agentic Workflow Audit Agent ▅▅▅▆▆▅▆▇▇█  latest=345.0 avg=289.6
CLI Version Checker          █▆▆▇▆▆▆▆▆▆  latest=131.5 avg=141.5
Tidy                         ▇▇█▆▆▅▆▆▆▇  latest=80.4 avg=48.6
Smoke Copilot                █▇▇▇▇▆▆▆▆▆  latest=43.0 avg=64.1

Chart 2 — Weekly Forecast Distribution (P10 optimistic / P50 median / P90 conservative), top 10 workflows

Weekly Forecast Distribution (P10 optimistic / P50 median / P90 conservative)

Go Logger Enhancement      [ggggggbbbbbbbbbbrrrrrrrrrrrrrr] P50=1729
Agentic Workflow Audit Age [ggggggbbbbbbbbrrrrrrrrrrrr----] P50=1569
CLI Version Checker        [gggbbbbrrrrrr-----------------] P50=831
Tidy                       [ggbbrrr-----------------------] P50=463
Copilot Agent PR Analysis  [gbbbrrr-----------------------] P50=443
Lockfile Statistics Analys [gbbrr-------------------------] P50=351
Dev                        [gbbrr-------------------------] P50=326
Duplicate Code Detector    [gbrrr-------------------------] P50=304
Smoke Copilot              [gbrr--------------------------] P50=272
GitHub MCP Remote Server T [bbrr--------------------------] P50=238

Legend: g=P10(opt) b=P50(median) r=P90(conservative)

Workflow Table (active workflows, sorted by observed AIC)

Workflow	Samples	Observed AIC	P50/run	P95/run	Weekly P50	Monthly P50	Success Rate	Monthly P10–P90
Go Logger Enhancement	28	10,389.21	365.67	678.34	1,729.08	7,762.13	75.0%	5,015.82 – 11,049.74
Agentic Workflow Audit Agent	27	7,818.27	288.39	408.03	1,569.46	6,944.84	88.9%	4,577.07 – 9,696.07
CLI Version Checker	27	3,821.09	132.65	206.86	830.55	3,675.83	96.3%	2,402.03 – 5,141.29
Tidy	47	2,284.20	31.07	127.02	462.67	2,039.59	89.4%	1,438.78 – 2,723.14
Smoke Copilot	34	2,180.38	58.37	93.77	271.63	1,218.63	55.9%	794.83 – 1,712.49
Copilot Agent PR Analysis	24	2,066.69	78.34	170.82	443.20	1,980.53	95.8%	1,276.23 – 2,821.24
Lockfile Statistics Analysis Agent	27	1,616.84	55.94	89.23	350.61	1,563.66	96.3%	1,038.99 – 2,170.67
Duplicate Code Detector	28	1,457.34	30.50	141.45	304.44	1,402.38	96.4%	865.41 – 2,046.55
Dev	29	1,448.56	23.21	101.83	326.10	1,446.31	100.0%	942.67 – 2,040.41
Terminal Stylist	31	962.87	27.52	49.32	219.32	969.07	100.0%	669.62 – 1,310.95
Daily News	23	879.69	32.62	83.54	198.53	886.25	100.0%	568.73 – 1,266.37
Smoke Claude	10	748.45	68.39	105.57	160.36	739.05	100.0%	375.37 – 1,256.85
GitHub MCP Remote Server Tools Report Generator ⚠️	3	731.37	244.29	248.69	238.39	731.37	100.0%	238.39 – 1,712.93
Weekly Workflow Analysis ⚠️	3	439.55	137.68	214.72	87.15	439.55	100.0%	87.15 – 1,057.22
Daily Documentation Updater	17	424.05	23.55	34.22	96.27	428.55	100.0%	251.68 – 635.32
Documentation Unbloat	30	409.27	13.20	17.66	90.20	397.35	96.7%	274.31 – 541.07
Smoke Codex ⚠️	9	323.14	41.73	56.03	58.78	286.76	88.9%	125.17 – 501.62
Weekly Issue Summary ⚠️	4	288.90	67.23	113.46	39.74	215.18	75.0%	39.74 – 491.83
Scout ⚠️	3	236.32	85.77	91.31	59.24	236.32	100.0%	59.24 – 563.96
Artifacts Usage Report ⚠️	4	159.80	39.51	46.55	39.51	162.57	100.0%	44.41 – 346.78
Repository Tree Map Generator ⚠️	4	79.60	17.39	26.09	17.36	60.80	75.0%	17.36 – 131.78
Smoke OpenCode ⚠️	1	22.91	22.91	22.91	0.00	22.91	100.0%	0.00 – 91.63
Plan Command ⚠️	1	19.59	19.59	19.59	0.00	19.59	100.0%	0.00 – 78.37

⚠️ = Monte Carlo projection flagged is_reliable: false (too few samples — see Data Quality section).

Data Quality & Accuracy

• 26 workflows had sampled_runs = 0 in the 30-day window (e.g. CI, CodeQL, Dependabot Updates, Test, CI Failure Doctor, and 21 others). These are excluded from AIC totals and projections; they are either dormant, disabled, or trigger-gated workflows that did not run recently. This is expected behavior, not a data defect, but it means the forecast covers only currently-active workflows — reactivating any excluded workflow would add unbudgeted spend.
• 9 active workflows have unreliable Monte Carlo projections (is_reliable: false) due to sparse samples (1–9 runs in 30 days): GitHub MCP Remote Server Tools Report Generator (n=3), Weekly Workflow Analysis (n=3), Smoke Codex (n=9), Weekly Issue Summary (n=4), Scout (n=3), Artifacts Usage Report (n=4), Repository Tree Map Generator (n=4), Smoke OpenCode (n=1), Plan Command (n=1). Their P10/P50/P90 spreads are wide relative to the median and should be treated as low-confidence placeholders, not budget commitments.
• Date windows are consistent: all sampled workflows show a ~30-day span (2026-07-22 to 2026-08-21), matching the requested history window — no stale or truncated sampling was detected.
• No zero/missing AIC values were found among active workflows — every workflow with sampled_runs > 0 has a non-zero avg_aic, p50_aic_per_run, and p95_aic_per_run.
• No outlier run frequencies detected: sampled run counts per workflow (1–47) are plausible given each workflow's trigger cadence (daily/weekly cron vs. manual/dispatch-only workflows).
• Follow-up performed: the prepared gh aw forecast output was validated by cross-checking workflow-level run_samples sums against reported avg_aic/projections; totals reconciled within rounding. No re-run of gh aw forecast --eval was required since the sampled data was internally consistent and sufficiently populated for the top-spending workflows that dominate the budget.

Zero-sample (inactive) workflows excluded from forecast

Mergefest, .github/workflows/test-proxy, Doc Build - Deploy, Notion Issue Summary, Commit Changes Analyzer, Poem Bot - A Creative Agentic Workflow, Q, Rebuild the documentation after making changes, Go Pattern Detector, Resource Summarizer Agent, Copilot Setup Steps, Sentry Issue Analyzer, CodeQL, MCP Inspector Agent, CI Failure Doctor, Dependabot Updates, CI, Test, Test Claude, Test Copilot CLI Engine, Test Copilot GitHub Integration, Basic Research Agent, Video Analysis Agent, Format, Lint, Build and Commit, Dev Hawk, copilot only

Full active-workflow raw metrics

Workflow	Sampled Runs	Success Rate	Avg AIC	P50 AIC	P95 AIC	Monthly Reliable
Tidy	47	89.4%	48.60	31.07	127.02	✅
Smoke Copilot	34	55.9%	64.13	58.37	93.77	✅
Terminal Stylist	31	100.0%	31.06	27.52	49.32	✅
Documentation Unbloat	30	96.7%	13.64	13.20	17.66	✅
Dev	29	100.0%	49.95	23.21	101.83	✅
Go Logger Enhancement	28	75.0%	371.04	365.67	678.34	✅
Duplicate Code Detector	28	96.4%	52.05	30.50	141.45	✅
Agentic Workflow Audit Agent	27	88.9%	289.56	288.39	408.03	✅
CLI Version Checker	27	96.3%	141.52	132.65	206.86	✅
Lockfile Statistics Analysis Agent	27	96.3%	59.88	55.94	89.23	✅
Copilot Agent PR Analysis	24	95.8%	86.11	78.34	170.82	✅
Daily News	23	100.0%	38.25	32.62	83.54	✅
Daily Documentation Updater	17	100.0%	24.94	23.55	34.22	✅
Smoke Claude	10	100.0%	74.84	68.39	105.57	✅
Smoke Codex	9	88.9%	35.90	41.73	56.03	⚠️ no/low
Weekly Issue Summary	4	75.0%	72.22	67.23	113.46	⚠️ no/low
Artifacts Usage Report	4	100.0%	39.95	39.51	46.55	⚠️ no/low
Repository Tree Map Generator	4	75.0%	19.90	17.39	26.09	⚠️ no/low
GitHub MCP Remote Server Tools Report Generator	3	100.0%	243.79	244.29	248.69	⚠️ no/low
Weekly Workflow Analysis	3	100.0%	146.52	137.68	214.72	⚠️ no/low
Scout	3	100.0%	78.77	85.77	91.31	⚠️ no/low
Smoke OpenCode	1	100.0%	22.91	22.91	22.91	⚠️ no/low
Plan Command	1	100.0%	19.59	19.59	19.59	⚠️ no/low

Assumptions & Methodology

• AIC (Actions-Incurred Cost / "AI Compute" proxy) figures are taken directly from gh aw forecast output (run_samples[].aic), summed per workflow for the observed total.
• Weekly/Monthly P10/P50/P90 projections come from the tool's built-in Monte Carlo simulation (10,000 iterations per workflow) over observed per-run AIC and run frequency in the 30-day history window.
• P10 (10th percentile) = optimistic scenario, P50 (50th percentile) = median/expected scenario, P90 (90th percentile) = conservative scenario, used consistently throughout this report.
• Workflows with 0 sampled runs are treated as currently inactive and are not included in the aggregate totals; no AIC values were invented for them.

Generated by 📈 Daily Spending Forecast · auto · 62.7 AIC · ⌖ 7.04 AIC · ⊞ 11.1K · ◷

• expires on Aug 28, 2026, 2:11 AM UTC-08:00