[lockfile-stats] Lockfile Statistics Report — 2026-10-12 (338 workflows)

[github-actions[bot]]

Lockfile Statistics — 2026-10-12

Analysis of 338 compiled .github/workflows/*.lock.yml files (54.72 MB total), single-script compact JSON pass. 0 malformed/skipped. Compared against prior snapshot 2026-10-10 (334 lockfiles).

File size distribution

Metric	Value
Count	338
Total bytes	54,720,747 (54.7 MB)
Avg	161,896 B (158.1 KB)
Min	90,534 B
Max	290,721 B

Trigger analysis

Trigger	Count	Δ vs 10-10
workflow_dispatch	329	+4
schedule	245	+4
pull_request	47	+4
workflow_call	5	0
issues	5	0
issue_comment	3	0
push	2	0
discussion / discussion_comment / pull_request_review_comment / workflow_run / deployment_status	1 each	0

Top combinations: schedule+workflow_dispatch (208), workflow_dispatch alone (71), pull_request+schedule+workflow_dispatch (33, +4). Dominant cron: 0 0 */2 * * (54 workflows, +4).

Safe outputs analysis

• create_discussion_workflows: 93 (unchanged); category resolved for 92 (99%). 1 unresolved: pr-sous-chef.lock.yml.
• safe_outputs_config_missing: 1 lockfile has no detectable GH_AW_SAFE_OUTPUTS_CONFIG.
• Discussion categories: audits 79, announcements 5, artifacts 2, dev 2, research 2, general 1, daily-news 1.
• Top safe-output types: create_issue 177 (+4), create_discussion 92, add_comment 74, create_pull_request 65, push_repo_memory 33, add_labels 32, upload_asset 26.
• Baseline tools (noop/missing_data/missing_tool/report_incomplete) present in 331 workflows (+4), consistent with near-universal safe-outputs scaffolding.

Structural characteristics

Metric	Total	Max	Δ Total
Jobs	2,404	23	+24
Steps	49,069	326	+839
Script steps	21,652	97	+217

Timeout-minutes distribution (per job): 10min → 382 (+4), 45min → 335 (+4), 60min → 327 (+4); long tail at 90/120/180/300.

Permission patterns (agent job)

Scope	read	write	none
contents	338	0	0
issues	261	0	0
pull-requests	255	0	0
actions	140	0	0
discussions	53	0	0
security-events	14	0	0
copilot-requests	0	162	0
id-token	0	2	0
models / packages / vulnerability-alerts / deployments / drives	1 each	0	0

permissions_unknown: 0/338. The agent job itself grants zero write scopes anywhere except copilot-requests/id-token.

Union across all jobs per workflow tells a different story: 336/338 (99.4%) grant some write scope somewhere (issues 336, contents 228, copilot-requests 162, pull-requests 150, discussions 96, actions 142, checks 14).

Engine distribution

Engine	Count	%	Δ
copilot	138	40.8%	+2
codex	76	22.5%	0
claude	62	18.3%	+1
pi	36	10.7%	+1
aider / goose	4 each	—	0
agy / opencode	3 each	—	0
crush / cursor / deepseek-harness / gemini / kiro / pydantic-ai	2 each	—	0

engine_unknown: 0/338 — every lockfile resolved via gh-aw-metadata.agent_id.

Tool & MCP patterns

Server	Workflows
safeoutputs	314 (+4)
github	189 (+4)
work-queue	33
agenticworkflows	27
mcpscripts	25
serena	17
agy-native / tavily	3 each
sentry	2

7 lockfiles required legacy allowed-tool-comment fallback parsing (no mcp_servers manifest).

Interesting findings

1. Templated batch growth: every trigger/safe-output/job/step/engine counter that changed since 2026-10-10 moved by almost exactly +4, matching the 4 new lockfiles added — strongly suggests the 4 new workflows share a near-identical template (same triggers, same safe-output set, same timeout profile) rather than being independently authored.
2. Read-only agent, write-elsewhere pattern is near-universal: the agent job itself holds 0 write-scope grants outside copilot-requests/id-token, yet 99.4% of workflows grant some write scope when permissions are unioned across all jobs — confirming safe-outputs' split-job write isolation is the dominant security pattern in this repo, not an exception.
3. copilot-requests is write-only: 162 workflows grant copilot-requests: write and none grant it read — an unusual scope that doesn't follow the typical read/write duality of other scopes.
4. Engine mix skews away from Claude: despite this being the gh-aw tooling repo, copilot (40.8%) and codex (22.5%) combined outnumber claude (18.3%) nearly 3.5:1.
5. Discussion category coverage gap: pr-sous-chef.lock.yml is the one workflow with create_discussion configured but no resolvable category, and a separate single lockfile is entirely missing safe-outputs config — both unchanged from 2026-10-10, i.e. known, persistent gaps rather than new regressions.

Historical trends (vs 2026-10-10)

Metric	10-10	10-12	Δ
Lockfiles	334	338	+4
Total bytes	54,061,748	54,720,747	+658,999
create_issue	173	177	+4
create_discussion	92	92	0
Jobs (total)	2,380	2,404	+24
Steps (total)	48,230	49,069	+839
workflows_with_any_write	332	336	+4
copilot engine	136	138	+2
claude engine	61	62	+1

Recommendations

• Confirm the 4 newly added workflows are intentionally templated; if so, consider documenting the shared template to avoid drift.
• Resolve the pr-sous-chef.lock.yml discussion-category gap and confirm whether the 1 lockfile missing GH_AW_SAFE_OUTPUTS_CONFIG is intentionally output-free.
• Audit the 162 copilot-requests: write-only grants to confirm no workflow needs read access instead/also.
• Continue monitoring the agent-job-read-only / union-job-write split as the baseline security posture; flag any new workflow where the agent job itself (not a downstream job) requests a write scope.

Methodology note: single-script compact JSON analysis (cached analyzer, reused across runs; one pass over all lockfiles, aggregate-only reasoning afterward).

Generated by 📊 Lockfile Statistics Analysis Agent · claude · agent · 73.7 AIC · ⊞ 6.5K · ◷

• expires on Oct 13, 2026, 12:31 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Lockfile Statistics Analysis Agent.

A newer discussion is available at Discussion #68530.