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

[github-actions]

Overview

This report analyzes GitHub Actions artifact usage across all 30 workflows in github/gh-aw. Data was collected from recent runs (up to the last 90 days). Of the 30 workflows, 10 generate artifacts while the remaining 20 produce none (they act as orchestration/automation triggers).

The repository generates approximately ~2.6 GB of active artifact storage at any given time, dominated by the CI pipeline and Doc Build workflows.

---

Summary Table

Workflow	Total Runs (All-Time)	Artifacts/Run	Size/Run	Est. Active Storage¹	Latest Run	Status
🔴 CI	17,808	31	~10.5 MB	~1.2 GB	2026-02-22	Active
🔴 Doc Build - Deploy	3,934	1	~212 MB	~777 MB	2026-02-22	Active
🟠 Smoke Codex	1,983	6	~2.6 MB	~434 MB	2026-02-22	Active
🟡 Smoke Claude	2,007	7	~486 KB	~80 MB	2026-02-22	Active
🟡 Smoke Copilot	2,071	8	~295 KB	~50 MB	2026-02-22	Active
🟢 Agentic Workflow Audit	213	10	~1.7 MB	~31 MB	2026-02-22	Active
🟢 Dev	1,637	7	~180 KB	~24 MB	2026-02-21	Active
🟢 CLI Version Checker	326	7	~199 KB	~5 MB	2026-02-21	Active
🟢 Artifacts Summary	85	5	~223 KB	~3 MB	2026-02-15	Active
🟢 Terminal Stylist	67	7	~131 KB	~1 MB	2026-02-21	Active
⚪ Scout	23,468	0	—	—	2026-02-22	Active
⚪ Plan Command	11,812	0	—	—	2026-02-22	Active
⚪ Mergefest	10,638	0	—	—	2026-02-22	Active
⚪ Copilot coding agent	13,207	0	—	—	2026-02-22	Active
⚪ CI Failure Doctor	2,670	0	—	—	2026-02-22	Active

¹ Active storage estimate = monthly run rate × size/run × retention window / 30 days. Artifact retention varies: Doc Build has 1-day retention; CI test results ~7 days; agentic workflows ~90 days.

---

Storage Breakdown by Category

Category	Workflows	Est. Active Storage
CI & Build	CI, Doc Build - Deploy	~2.0 GB
Smoke / Regression Tests	Smoke Codex, Claude, Copilot	~564 MB
Agentic Workflows	Dev, CLI Version Checker, Audit, Terminal Stylist	~61 MB
Orchestration (no artifacts)	Scout, Plan, Mergefest, Copilot agent, CI Doctor, etc.	0
Total		~2.6 GB

---

Artifact Types Breakdown (per workflow category)

CI Pipeline Artifacts (31 per run)

Artifact Name	Purpose	Retention
test-result-unit	Unit test JUnit results	~7 days
test-result-integration-* (20+)	Per-suite integration test results	~7 days
gh-aw-linux-amd64	Built binary	~7 days
coverage-report	HTML coverage report	~7 days
test-coverage-analysis	Coverage JSON data	~7 days
benchmark-results	Go benchmark output	~7 days
safe-outputs-conformance-report	Conformance check	~7 days
dependency-audit	Dependency security audit	~7 days

Agentic Workflow Artifacts (7–10 per run)

Artifact Name	Purpose	Retention
prompt	Input prompt to the agent	1 day
safe-output	Compressed safe output	90 days
agent-output	Raw agent response	90 days
agent_outputs	Full output bundle	90 days
agent-artifacts	Agent-generated files	90 days
cache-memory	Persistent memory across runs	90 days
repo-memory-default	Repo-scoped memory	1 day
safe-outputs-assets	Images/attachments from safe outputs	1 day
threat-detection.log	Security scan log	90 days
safe-output-items	Structured output items	90 days
*.dockerbuild	Docker build cache metadata	90 days

High-Volume Workflows (No Artifacts)

These workflows run very frequently but produce zero artifacts — they are pure automation triggers or approval gates:

Workflow	All-Time Runs	Role
Scout	23,468	PR review comment automation
Copilot coding agent	13,207	SWE agent task dispatch
Plan Command	11,812	Issue planning automation
Mergefest	10,638	PR merge automation
CI Failure Doctor	2,670	CI error analysis (triggers on failures)
copilot only	96	Manual Copilot trigger

---

Key Insights

🔴 Biggest Storage Consumers

1. CI workflow — 31 artifacts per run, ~10.5 MB/run, running ~500×/month → ~1.2 GB active. The test matrix generates 20+ test-result-integration-* artifacts per run.
2. Doc Build - Deploy — Single large ~212 MB github-pages artifact, but 1-day retention keeps active storage to ~777 MB despite 3,934 total runs.
3. Smoke Codex — Surprisingly large agent-artifacts artifact (~2.5 MB/run), significantly larger than Claude or Copilot equivalents.

📈 Trends

• Scout (23,468 runs), Plan Command (11,812 runs), and Mergefest (10,638 runs) are the most-run workflows but generate no storage overhead.
• The agent-artifacts size varies significantly by engine: Codex (~2.5 MB) >> Claude (~437 KB) >> Copilot (~191 KB), suggesting Codex generates more verbose output files.
• Agentic Workflow Audit Agent uses the richest artifact set (10 types including cache-memory, repo-memory-default, safe-outputs-assets) — this is by design for its cross-run memory feature.

---

Recommendations

Priority	Recommendation	Impact
🔴 High	Reduce CI artifact count — Consolidate the 20+ test-result-integration-* artifacts into a single report artifact to reduce per-run count from 31 → ~10	~60% reduction in CI artifact count
🟠 Medium	Investigate Smoke Codex agent-artifacts size — At 2.5 MB/run vs 191 KB for Copilot, Codex generates ~13× larger artifacts. Trimming verbose agent output could save ~400 MB of active storage	~400 MB saved
🟡 Low	Review prompt artifact retention — Currently 1 day, which is appropriate. Keep as-is.	No change
🟢 Low	CI test result retention — Currently ~7 days, appropriate for debugging failures. Could reduce to 3 days if failures are addressed quickly	Minor CI storage savings

---

References:

• §22271526407 — This workflow run
• §22030637592 — Artifacts Summary (2026-02-15)
• §22271394279 — CI (2026-02-22)

---

Note: This was intended to be a discussion, but discussions could not be created due to permissions issues. This issue was created as a fallback.

Tip: Discussion creation may fail if the specified category is not announcement-capable. Consider using the "Announcements" category or another announcement-capable category in your workflow configuration.

Generated by Artifacts Summary

• expires on Mar 1, 2026, 5:58 AM UTC