Artifacts Usage Report — github/gh-aw (April 2026) #24677
Replies: 3 comments
|
🤖 The smoke test agent was here! Beep boop — testing, testing, 1-2-3. All systems nominal! 🚀
|
|
💥 WHOOSH! The smoke test agent has ARRIVED, blazing through the cosmos like a comet with a clipart flame! 🔥 ⚡ POW! — Claude engine 23998581834 swooped in, ran its tests, and declared this repository MOSTLY NOMINAL! 🦸 With the power of 16 passing tests and only 1 rate-limited Tavily villain standing in the way, the smoke test hero lives to fight another day! 💫 WHAM! — Safe outputs? DEPLOYED! PR reviews? SUBMITTED! Slack messages? SENT! Code scanning alerts? FIRED! To be continued... in the next workflow run! 🚀
|
|
This discussion was automatically closed because it expired on 2026-04-06T06:16:26.734Z.
|
Uh oh!
There was an error while loading. Please reload this page.
Analysis of GitHub Actions artifact usage across the 241 workflows in this repository, based on data collected from recent runs (last 30 days, ending 2026-04-05).
Summary
.lock.ymlworkflows)github-pages– 234 MB per run (Doc Build)Artifact Usage by Workflow
* Runs marked skipped or triggered conditionally — no artifacts generated.
** Run failed before producing artifacts.
Standard Agentic Workflow Artifact Pattern
All
.lock.ymlagentic workflows share a consistent artifact schema:activationagentdetectionsafe-output-itemsfirewall-audit-logsrepo-memory-default*.dockerbuildapmcode-scanning-sarifCI Workflow Artifacts (37 per run)
The CI workflow is the most complex, generating test result artifacts for every job matrix combination:
View all 37 CI artifact types
test-result-unitcoverage-reporttest-coverage-analysisbenchmark-resultsfuzz-results-Workflow-Triggersfuzz-results-Workflow-Corefuzz-results-Workflow-Parsingfuzz-results-Parsertest-result-integration-CLI MCP Connectivitytest-result-integration-CLI Docker Buildtest-result-integration-CLI Progress Flagtest-result-integration-CLI HTTP MCP Connecttest-result-integration-CLI Completion & Othertest-result-integration-CLI MCP Othertest-result-integration-CLI Compile & Poutinetest-result-integration-Parser Remote Fetch & Cachetest-result-integration-Workflow Rendering & Bundlingtest-result-integration-Workflow Compilertest-result-integration-CLI Security Toolstest-result-integration-Workflow Misc Part 2test-result-integration-CLI Compile Workflowstest-result-integration-Workflow Actions & Containerstest-result-integration-add-dispatch-workflowtest-result-integration-Workflow Tools & MCPtest-result-integration-CLI Add & List Commandstest-result-integration-CMD Teststest-result-integration-Workflow Featurestest-result-integration-CLI MCP Inspect GitHubtest-result-integration-unauthenticatedtest-result-integration-Workflow ValidationTotal per CI run: ~3.0 MB
Storage Estimates (30-Day Window)
* Doc Build generates 234 MB artifacts but expires them in 1 day, so at any moment only ~20 GB of doc artifacts are live.
Insights & Recommendations
🔴 High Priority
Doc Build artifact size — The
github-pagesartifact is 234 MB per run and the workflow has run 7,060+ times. The 1-day retention policy appropriately limits accumulation, but the per-run size warrants investigation — consider whether the build can be optimized (e.g., compression, incremental builds) to reduce from ~234 MB.Smoke Codex
agentartifact — At ~3.4 MB per run (vs ~0.35 MB for Smoke Copilot and ~0.57 MB for Smoke Claude), the Codex agent artifact is ~6-10× larger. This suggests verbose output or uncompressed logs — worth investigating.🟡 Medium Priority
CI artifact volume — With 37 artifacts per run and ~1,800 runs/month, CI generates ~66,000+ artifacts/month. The 14-day retention is reasonable, but consolidating the many small
test-result-integration-*artifacts into a single archive could reduce artifact count significantly.repo-memory-defaultin Daily workflows — Daily workflows (e.g., Daily News, Daily Repository Chronicle) upload ~1.1 MBrepo-memory-defaultartifacts each run but expire them after only 1 day. If these represent cache-memory state that should persist between daily runs, a longer retention (e.g., 7 days) might improve reliability.Failed workflow artifacts — Several daily workflows (Daily Fact, Daily News) are failing before producing artifacts. This suggests runtime issues that prevent artifact upload, not just content failures. These should be investigated.
🟢 Already Optimized
firewall-audit-logs,detection) with 90-day retention — Appropriate for security audit trails.Methodology
Data collected via GitHub Actions API for all 241 workflows in
github/gh-aw. Artifact statistics are based on the most recent completed (non-cancelled) run per workflow. Storage estimates use per-run sizes multiplied by estimated monthly run frequencies derived from total run counts and workflow creation dates.References:
All reactions