Artifacts Usage Report — github/gh-aw (March 2026)

[github-actions[bot]]

Overview

This report summarizes GitHub Actions artifact usage across the 229 workflows in github/gh-aw, based on analysis of recent runs (March 2026). The repository is highly active (~94 CI runs/day) with a diverse mix of CI infrastructure workflows and agentic workflows. Artifact production is dominated by two workflows: CI (high artifact count) and Doc Build - Deploy (largest single artifact by size).

---

Artifact Summary Table

Metrics are based on analysis of the most recent completed runs per workflow. Estimated 30-day totals use observed daily rates since repo creation (Aug 2025 ≈ 210 days).

Workflow Name	Artifacts/Run	Size/Run	Retention	Est. Runs/Day	Est. 30-day Artifacts	Est. 30-day Data Created	Status
CI	32–37	2.6–11 MB¹	14 days	~94	~50,000–84,000	~4–20 GB	🟢 Very Active
Doc Build - Deploy	1	233 MB	1 day	~26	~780	~182 GB	🟢 Very Active
Smoke Codex	4	~2.7 MB	90 days	~11	~1,320	~3.6 GB	🟢 Active
Smoke Claude	6	~0.37 MB	90 days	~11	~1,980	~0.73 GB	🟢 Active
Smoke Copilot	5	~0.31 MB	90 days	~11	~1,650	~0.62 GB	🟢 Active
Dev	4	~0.28 MB	90 days	~8	~960	~0.27 GB	🟢 Active
CI Failure Doctor	0²	—	—	~17	~0	—	🟡 Conditional
Mergefest	0³	—	—	~60	~0	—	🟡 Mostly Skipped
Documentation Unbloat	0³	—	—	~65	~0	—	🟡 Mostly Skipped
All other workflows (~220)	varies	small	90 days	low	low	minimal	🔵 Low Volume

¹ Full CI runs (main branch merges) include the gh-aw-linux-amd64 binary (~8.35 MB); PR-only runs omit it.
² CI Failure Doctor only runs when CI fails; no artifacts on skipped runs.
³ These workflows skip on most triggers; no artifacts produced on skip.

---

Artifact Breakdown by Workflow

CI Workflow — Artifact Detail

Each CI run generates 32–37 artifacts. Notable artifacts per run:

Artifact Name	Typical Size	Notes
gh-aw-linux-amd64	~8.35 MB	Binary build, only on full runs
coverage-report	~1.20 MB	HTML coverage report
test-result-unit	~1.10 MB	Unit test results
test-coverage-analysis	~0.07 MB	Coverage analysis
test-result-integration-* (×~20)	0.01–0.05 MB each	One per integration test group
fuzz-results-* (×4)	<0.01 MB each	Fuzzing results
benchmark-results	<0.01 MB	Benchmark data

Retention: 14 days (default). With ~94 runs/day, approximately ~1,300 active runs at any time × ~35 artifacts = ~45,500 live artifacts from CI alone.

Doc Build - Deploy — Artifact Detail

Artifact Name	Typical Size	Retention
github-pages	~233 MB	1 day

This workflow triggers on every push to main alongside CI. The github-pages artifact is a full Astro Starlight build. The 1-day retention means storage stays bounded (~26 × 233 MB ≈ 6 GB at any given time), but ~182 GB of data is created and discarded each month.

Smoke Tests — Artifact Detail

Smoke tests (Smoke Claude, Smoke Codex, Smoke Copilot) each generate a standard set of agentic workflow artifacts per run:

Artifact Name	Typical Size	Retention	Notes
agent	0.27–2.6 MB	90 days	Agent logs (varies significantly by engine)
detection	0.003–0.096 MB	90 days	Threat detection output
github~gh-aw~*.dockerbuild	~24–25 KB	90 days	Docker build logs (Claude/Copilot only)
safe-output-items	<1 KB	90 days	Safe output metadata
activation	~2–8 KB	1 day	Activation job logs
apm	~1 KB	1 day	APM data

Notable: Smoke Codex has the largest agent artifact (~2.6 MB vs ~0.3 MB for others), suggesting more verbose logging.

---

Insights & Recommendations

🏆 Highest Artifact Count

CI generates by far the most artifacts (~35/run × ~94 runs/day ≈ ~3,290 new artifacts/day). With 14-day retention, approximately 46,000+ live artifacts exist at any time from CI alone.

📦 Highest Data Volume Created

Doc Build - Deploy creates the most data (~233 MB/run × ~26 runs/day ≈ ~6 GB/day), but 1-day retention keeps live storage minimal. Over 30 days, ~182 GB is generated and discarded.

💡 Optimization Opportunities

1. Consolidate CI test artifacts: The ~20 separate test-result-integration-* artifacts (each ~10–50 KB) could be merged into a single test-results artifact, reducing artifact count by ~55% per run with minimal workflow changes.

2. Review binary artifact retention: The gh-aw-linux-amd64 binary (~8.35 MB, 14-day retention) is generated on main-branch runs. Reducing retention to 7 days or limiting to tagged releases could save significant storage.

3. Smoke Codex agent log size: At ~2.6 MB vs ~0.3 MB for other engines, Codex agent logs are ~8× larger. Review log verbosity settings.

4. 90-day retention on smoke test artifacts: With 3 smoke workflows running ~11×/day each, 90-day retention means ~29,700 live smoke test artifacts at any time. Reducing to 30 days would cut this by ~67%.

5. Large number of low-volume agentic workflows: The remaining ~220 workflows each run infrequently (schedule/dispatch) with small artifacts. These contribute minimally to storage individually but collectively may add up.

---

Workflow Activity Distribution

Run Volume by Workflow (top 10, all time)

Workflow	Total Runs	Type
CI	19,723	CI/CD
Documentation Unbloat	13,597	Agentic
Mergefest	12,552	Agentic
Doc Build - Deploy	5,534	CI/CD
CI Failure Doctor	3,539	Agentic
Smoke Copilot	2,382	Smoke Test
Smoke Claude	2,281	Smoke Test
Smoke Codex	2,253	Smoke Test
Dev	1,663	Agentic
Daily News	151	Agentic

Note: Most Mergefest/Documentation Unbloat runs are skipped and produce zero artifacts.

---

References:

• §23103378028 — Latest CI run (success)
• §23103378032 — Latest Doc Build run (success)
• §23102624714 — Latest Smoke Claude run (success)

AI generated by Artifacts Summary · history

• expires on Mar 16, 2026, 6:02 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-03-16T06:02:06.338Z.

Closed by Workflow