[lockfile-stats] Lockfile Statistics Audit — 2026-07-14 (256 workflows, ~30 MB) #45552
Closed
Replies: 1 comment
|
This discussion has been marked as outdated by Lockfile Statistics Analysis Agent. A newer discussion is available at Discussion #45816. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Executive Summary
Analysis of 256 compiled
.github/workflows/*.lock.ymlfiles on 2026-07-14. All parsed cleanly withyaml.safe_load— 0 skipped, 0 malformed.example-permissions-warning)smoke-copilot)The corpus is large and remarkably uniform: compiled scaffolding (avg 117.6 steps/workflow) dominates each file, so lockfile size clusters tightly around ~120 KB regardless of agent logic.
File Size Distribution
Largest & smallest
Largest:
smoke-copilot(203 KB),smoke-copilot-aoai-entra(190 KB),smoke-copilot-aoai-apikey(190 KB),smoke-claude(184 KB),smoke-copilot-arm(175 KB),smoke-codex(165 KB),mcp-inspector(157 KB),deep-report(154 KB).Smallest:
example-permissions-warning(81 KB),firewall(82 KB),codex-github-remote-mcp-test(83 KB),hippo-embed(90 KB),ace-editor(90 KB).Trigger Analysis
workflow_dispatchschedulepull_requestissuespush/issue_comment/workflow_callTop combinations:
schedule+workflow_dispatch(171),workflow_dispatchonly (44),pull_request+workflow_dispatch(28).175 cron schedules are well jittered — minute offsets spread across the hour (e.g.
49 14,23 11,38 3) rather than clustering on:00, avoiding thundering-herd load. Cadence is mostly daily (* * *) with a weekday band (* * 1-5) and a few*/6hhigh-frequency jobs.Safe Outputs Analysis
Five tools appear in 251/256 workflows — the built-in safety scaffold (
noop,missing-tool,missing-data,report-incomplete,create-report-incomplete-issue). Excluding that baseline, the user-facing outputs:Discussion categories:
audits(76),announcements(5),artifacts(2),dev(2),daily-news(1),research(1).Structural Characteristics
release)smoke-copilot)Per-job timeouts: 256 jobs in 31–60 min band, 28 in 6–15 min, 2 at ≤5 min.
Permission Patterns
actions:readandcontents:readare universal (256).issues:write(249) is nearly universal — consistent with the create-issue/incomplete-issue safety pattern.Tool & MCP Patterns
MCP servers by workflow count:
github(100),serena(24),tavily(5),sentry(5),grafana(3),markitdown(3),brave-search(2),ast-grep(2),arxiv(2),deepwiki(2),notion(2),datadog(2), plus single-usesemgrep,agentdb,context7,memory.Engine distribution (regex-derived, approximate): copilot ≈64, claude ≈60, codex ≈14, plus one each antigravity/crush/gemini/opencode; a large
other/unclassified bucket (~114) reflectsengine:-line matches inside nested config. Thecopilot-requests:writepermission (67 workflows) is a cleaner proxy for Copilot-engine adoption.Interesting Findings
workflow_dispatchand 68%schedule;schedule+workflow_dispatchalone covers 171 workflows (67%). These are overwhelmingly cron-driven agents with a manual escape hatch, not event-reactive bots.noop/missing-tool/report-incompleteguardrail set — a near-invariant convention that any audit can rely on.serena(24) is a distant but clear second, signalling heavy code-navigation usage across the fleet.Historical Trends
Comparable v1 snapshots exist since 2026-07-10.
Day-over-day: −1 lockfile, −57 KB total — essentially flat.
Long-run (since 2026-05-20): 233 → 256 lockfiles (+9.9%) and 22.4 MB → 31.4 MB (+40%). Byte growth has far outpaced file count, confirming the scaffolding-per-file expansion noted in finding 3.
Recommendations
noop/missing-*/report-incompletebundle is valuable, but confirm every workflow genuinely needs the full set rather than inheriting it by default.issues:write(249) andcontents:write(130) are broad; workflows whose only output is a discussion or comment may not need issue/contents write.Methodology
Single-script compact JSON analysis: one cached Python analyzer (
lockfile_stats_v1.py) parsed all 256 lockfiles withyaml.safe_load(yaml_available: true) in a single pass, emitting a ~7 KB summary JSON; all figures above derive from that summary. Engine attribution and a few safe-output/category signals use regex fallback and are labelled approximate where noted.References: §29366173484
Warning
Firewall blocked 1 domain
The following domain was blocked by the firewall during workflow execution:
awmgmcpgSee Network Configuration for more information.
All reactions