[lockfile-stats] Lockfile Statistics — 2026-09-05 (299 workflows analyzed) #58866
Closed
Replies: 1 comment
|
This discussion has been marked as outdated by Lockfile Statistics Analysis Agent. A newer discussion is available at Discussion #59030. |
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.
Lockfile Statistics — 2026-09-05
Analysis of all
.github/workflows/*.lock.ymlcompiled agentic workflows ingithub/gh-aw. Data extracted via one-pass parser fromgh-aw-metadata/gh-aw-manifestcomments,GH_AW_SAFE_OUTPUTS_CONFIG, and per-job permissions blocks (299 lockfiles, 0 malformed/skipped).Executive summary
File size distribution
Trigger analysis
Top trigger combinations:
schedule+workflow_dispatch(209), bareworkflow_dispatch(38),pull_request+schedule+workflow_dispatch(29). Cron cadence is dominated by0 0 */2 * *(every-other-day) at 42 workflows; the remaining ~30 distinct crons are singletons or pairs, showing most scheduled workflows converge on the same twice-daily-ish slot rather than spreading load.Safe outputs analysis
92 workflows configure
create_discussion(category detection: 92/92 resolved, 0 unresolved, 0 fallback-parsed — clean). Discussion categories: audits 79, announcements 5, artifacts 2, dev 2, research 2, general 1, daily-news 1 — audits dominates, consistent with this workflow's own category.Most common non-builtin safe outputs:
create_issue(146),add_comment(77),create_pull_request(64),push_repo_memory(36, up from 34),add_labels(32).Structural characteristics
Timeout-minutes distribution (per job, across all workflows): 45min ×299, 60min ×292, 10min ×318, 90min ×3, 5min ×2, 120min ×2, 180min ×1, 15min ×1 — the 45/60/10 triad is the de facto standard across nearly every workflow.
Permission patterns (agent job)
contents: readis universal (299/299). Beyond that, most scopes default tonone:actionsnone 185 / read 114,discussionsnone 247 / read 52,security-eventsnone 285 / read 14. Onlyid-tokenshows anywrite(2 workflows). Union-across-jobs view shows where writes actually happen:issues: writeon all 299,contents: writeon 206,pull-requests: writeon 146,discussions: writeon 96 — i.e., every workflow's agent job itself is read-only/least-privilege, with writes isolated to dedicated safe-output jobs.Engine distribution
0 lockfiles had unresolved engine (
agent_idcleanly parsed fromgh-aw-metadatain all 299). Most common model pins:openai/gpt-5.4(44),copilot/mai-code-1-flash-picker(44),copilot/claude-sonnet-4.5(9).Tool & MCP patterns
safeoutputsMCP is present in all 299 workflows (expected — it's the safe-output transport). Beyond that:github189 (up from 121, +56%),agenticworkflows44,serena25,mcpscripts12,tavily5. Zero fallback parsing was needed (mcp_fallback_used_count: 0— all resolved viagh-aw-manifest).Within the
githubserver, tool adoption broadened meaningfully day-over-day: read counts for most tools rose from ~113-121 to ~171-189, and several tool categories appear as newly-adopted that weren't present yesterday, includingactions_get/actions_list/get_job_logs(17 workflows), andget_discussion/get_discussion_comments/list_discussion_categories/list_discussions(12 workflows).Interesting findings
piengine usage dropped 62% (29→11) whilecodexrose 24% (76→94) in a single day — consistent with a batch re-pin of workflows frompitocodex.contents: read-only with no direct write scopes beyond 2id-token: writeoutliers — all writes flow through dedicated safe-output jobs (unionissues: write= 299/299).create_discussionworkflows resolved a category with zero fallback-regex parsing needed — theGH_AW_SAFE_OUTPUTS_CONFIGJSON path covers 100% of cases in this corpus.Historical trends (vs 2026-09-04)
push_repo_memory+2,close_issue+1.Recommendations
pi→codexengine shift was intentional, consider documenting the rationale (cost/capability) so future audits don't flag it as drift.Methodology: single-script compact JSON analysis (
lockfile_stats_v4.py, cached in/tmp/gh-aw/cache-memory/scripts/). Engine resolved fromgh-aw-metadataagent_id; permissions from per-jobpermissions:blocks (top-levelpermissions: {}carries no signal); safe outputs and discussion categories fromGH_AW_SAFE_OUTPUTS_CONFIGJSON; MCP usage fromgh-aw-manifest.Warning
Firewall blocked 1 domain
The following domain was blocked by the firewall during workflow execution:
api.anthropic.comTo allow these domains, add them to the
network.allowedlist in your workflow frontmatter:See Network Configuration for more information.
All reactions