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

[github-actions[bot]]

This report summarizes GitHub Actions artifact usage across the github/gh-aw repository, based on analysis of 14 key workflows (out of 234 total) sampled from recent runs as of 2026-03-22.

---

Overview

The repository runs 234 workflows with a combined volume of well over 100,000 runs per month. Artifact generation is concentrated in a handful of workflows — most of the high-frequency workflows (Scout, Mergefest, Plan Command) generate no artifacts at all. The workflows that do generate artifacts fall into two tiers: large infrastructure workflows (CI, Doc Build) and recurring agentic workflows with cache/memory patterns.

---

Artifact Summary by Workflow

Workflow Name	Total Runs	Artifacts/Run	Size/Run	Latest Run	Notes
Doc Build - Deploy	6,092	1	222.80 MB	2026-03-22	github-pages artifact; 1-day retention
CI	20,436	37	11.76 MB	2026-03-22	Test results, coverage, binary; mixed retention
Smoke Codex	2,387	5	2.54 MB	2026-03-22	agent + qmd-index dominant
Daily Repository Chronicle	141	5	1.52 MB	2026-03-20	cache-memory 90d; safe-outputs 1d
Daily Firewall Logs Reporter	223	7	1.49 MB	2026-03-22	cache-memory 90d; dockerbuild included
Daily News	166	6	1.44 MB	2026-03-20	repo-memory-default 1d; agent 90d
Agentic Workflow Audit Agent	256	8	1.27 MB	2026-03-22	cache-memory 90d; 8 artifact types
Smoke Claude	2,410	6	0.48 MB	2026-03-22	agent + dockerbuild; 90d retention
Smoke Copilot	2,514	5	0.40 MB	2026-03-22	agent + dockerbuild; 90d retention
Dev	1,680	4	0.23 MB	2026-03-21	Minimal artifacts per run
Scout	28,459	0	N/A	2026-03-22	No artifacts generated
Mergefest	13,419	0	N/A	2026-03-22	No artifacts generated
Plan Command	14,671	0	N/A	2026-03-22	No artifacts generated
CI Failure Doctor	3,871	0	N/A	2026-03-22	Mostly skipped; no artifacts

ℹ️ Note: The analysis sampled the most recently active runs for each workflow. Artifact sizes represent non-expired artifacts at the time of sampling.

---

Estimated Total Storage (Active Artifacts)

Workflow	Retention	Active Runs (est.)	Est. Storage
Doc Build - Deploy	1 day	~10	~2.2 GB
CI	14–90 days	~150	~1.8 GB
Smoke Claude/Codex/Copilot	90 days	~270 (3×90)	~600 MB
Daily/Periodic agentic workflows	90 days	~450 (5×90)	~700 MB
Other agentic workflows	Mixed	~300	~400 MB
Total (estimated)			~5.7 GB

---

Insights

Top Storage Consumers

1. Doc Build - Deploy generates github-pages artifacts at ~223 MB each. With 10+ main-branch pushes per day and 1-day retention, this alone contributes ~2 GB of active storage at any given moment. Each artifact contains the full compiled static documentation site.

2. CI creates 37 artifacts per run — the largest contributor by artifact count. The binary artifact (gh-aw-linux-amd64, ~9 MB) and test/coverage reports (coverage-report, test-result-unit) dominate per-run size. With ~10 CI runs per day and 14-day retention for most artifacts, total CI storage is estimated at ~1.8 GB.

3. Smoke Codex generates larger artifacts than Claude/Copilot equivalents. The agent artifact reaches 1.4 MB and qmd-index adds another 1 MB — roughly 5× larger than Smoke Claude per run. Over 2,387 runs at 90-day retention, this contributes meaningfully to total storage.

Artifact Types and Patterns

Common artifact types identified across agentic workflows:

Artifact Name	Typical Size	Retention	Purpose
github-pages	223 MB	1 day	Built documentation site
gh-aw-linux-amd64	9 MB	14 days	Compiled binary
coverage-report	1.3 MB	14 days	HTML coverage report
test-result-unit	1.2 MB	14 days	Unit test results
cache-memory	0.9 MB	90 days	AI agent memory persistence
agent	0.2–1.4 MB	90 days	AI agent output/logs
safe-outputs-assets	0.3–0.4 MB	1 day	Safe output rendering assets
repo-memory-default	0.6–1 MB	1 day	Repository-level agent memory
detection	2–153 KB	90 days	Safe input detection result
activation	6–12 KB	1 day	Workflow activation metadata
safe-output-items	< 1 KB	90 days	Safe output items manifest
apm	< 1 KB	1 day	APM telemetry

Pattern observed: Agentic workflows produce two tiers of artifacts:

• Short-lived (1 day): activation, safe-outputs-assets, repo-memory-default, apm — ephemeral data only needed within the workflow run
• Long-lived (90 days): cache-memory, agent, detection, safe-output-items — persistent state and outputs

High-Volume Workflows Without Artifacts

Several of the highest-volume workflows generate zero artifacts:

Workflow	Total Runs	% of Total Runs
Scout	28,459	~23%
Plan Command	14,671	~12%
Mergefest	13,419	~11%
CI Failure Doctor	3,871	~3%

These workflows collectively account for ~49% of all workflow runs but contribute nothing to artifact storage — a positive efficiency indicator. Scout and Plan Command have the highest run volumes in the repository.

---

Recommendations

1. Review Doc Build retention policy — github-pages artifacts are automatically set to 1-day retention, which is appropriate. However, at ~223 MB per artifact and 10+ runs/day on main, consider whether all branch CI pushes trigger documentation builds or only main.

2. Consolidate CI test result artifacts — The CI workflow currently generates ~26 separate integration test result artifacts. These could be consolidated into a single test-results artifact using artifact merge, reducing artifact count from 37 to ~12 per run without losing data.

3. Audit cache-memory retention — Several agentic workflows store cache-memory artifacts for 90 days. Verify that workflows reading these artifacts can handle stale/missing data gracefully, and consider reducing to 30 days for workflows that run daily (the memory would be refreshed 90× within the retention window anyway).

4. Monitor Smoke Codex growth — qmd-index artifacts were not present in Claude/Copilot smoke runs, suggesting this is a newer feature. At ~1 MB per run × 90-day retention × daily runs = ~90 MB and growing, this warrants monitoring.

5. Add a 30-day rolling analysis — Run this artifact analysis workflow weekly to track storage trends and catch regressions before they accumulate.

---

References:

• §23396857124 — This workflow run
• CI Workflow
• Doc Build Workflow

AI generated by Artifacts Summary · history

• expires on Mar 23, 2026, 5:59 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-03-23T05:59:51.179Z.

Closed by Workflow