Workflow Scope Threshold Sample

Source issue: #898 Upstream audit: #884 / #877 Baseline input: workflow-metrics-baseline.md Sample window: PRs updated on or after 2026-04-24 with current scope-violation or scope-exception labels.

This note samples the recent PRs behind the #884 observation that scope-exception usage is about as frequent as scope-violation. It is intentionally docs-only: no Scope Police threshold, label, workflow, template, or auto-ready behavior changes are made here.

Method

Read-only commands:

gh pr list --repo nurockplayer/tachigo --state all --limit 120 \
  --search 'updated:>=2026-04-24 label:scope-violation,label:scope-exception' \
  --json number,title,state,author,labels,createdAt,updatedAt,closedAt,mergedAt,url,additions,deletions,changedFiles

gh pr list --repo nurockplayer/tachigo --state all --limit 100 \
  --search 'updated:>=2026-04-24 label:scope-exception' \
  --json number,title,state,author,labels,createdAt,updatedAt,closedAt,mergedAt,url,additions,deletions,changedFiles

for n in 862 844 827 825 730 729 556 554 476 457 434 372 369 342 \
  863 845 819 733 679 622 488 459 435 417 389; do
  gh pr view "$n" --repo nurockplayer/tachigo \
    --json number,title,state,additions,deletions,changedFiles,labels,files,mergedAt,closedAt,url
done

Important limitation: GitHub search exposes the labels currently attached to PRs, not a full historical label-event stream. The sample therefore counts current label occurrences. It found 14 scope-violation occurrences plus 15 scope-exception occurrences, with 4 overlapping PRs, for 25 unique PRs.

Summary

Classification	Unique PRs	Pattern
Legit-large / mechanical	13	Monorepo moves, generated Swagger, migration baseline, framework dependency rollout, large deletion, docs portal import, or one-shot governance extraction.
Splittable / maybe splittable	8	Mixed backend/frontend/schema work, broad frontend redesigns, config plus dependency cleanup, or dirty retry attempts.
False positive / artifact-heavy friction	4	Long docs, planning artifacts, source-small dependency churn, or binary-asset-heavy PRs where hard line limits caught reviewability but not product risk.

The result does not support raising the global 1000 hard or 600 soft line thresholds. Those limits caught several genuinely splittable PRs. The main calibration problem is that scope-exception is too undifferentiated: it covers legitimate generated/mechanical work, broad feature bundles, and docs-only planning artifacts with the same bypass semantics.

Sample

PR	Labels	Size	State	Classification	Notes
#342	scope-violation	+298/-40, 8 files	closed	Splittable	Backend and dashboard changes in one PR; the label correctly caught mixed product surfaces.
#369	scope-violation, scope-exception	+551/-814, 10 files	merged	Legit-large / mechanical	One governance extraction touching agent rules, hooks, CI, and metadata. Exception was plausible, but it should carry an explicit governance-extraction reason.
#372	scope-violation, scope-exception	+1691/-13, 5 files	closed	Legit-large / generated	Swagger-heavy backend rename slice. Generated-file handling needs a clearer category than generic exception.
#389	scope-exception	+1714/-5, 5 files	merged	Legit-large / generated	Swagger regeneration dominated the diff; exception was plausible with generated-doc evidence.
#417	scope-exception	+3262/-117, 151 files	merged	Legit-large / mechanical	Frontend monorepo move. This is the canonical mechanical-move exception class.
#434	scope-violation	+1011/-14, 49 files	closed	Legit-large / mechanical	Extension demo move; violation is expected without an explicit mechanical-move exception.
#435	scope-exception	+65/-36, 156 files	merged	Legit-large / mechanical	Backend service move; file-count exception was reasonable, but should be documented as path-only movement.
#457	scope-violation	+1346/-531, 19 files	closed	Splittable	Dashboard Refine migration included pages, providers, tests, docs, and lockfile. Better split provider foundation from page migration.
#459	scope-exception	+1101/-23, 8 files	merged	Legit-large / framework dependency	Dashboard provider foundation plus lockfiles. Exception was plausible, but needs dependency/artifact evidence.
#476	scope-violation, scope-exception	+2061/-21, 10 files	closed	Legit-large / migration	Atlas migration/tooling baseline. A migration-specific exception category would be clearer than general bypass.
#488	scope-exception	+786/-365, 9 files	merged	Legit-large / governance rollout	Auto-ready workflow rollout mixed CI, policy docs, agent docs, and plans. It was below the hard line limit but broad enough to justify stricter exception rationale.
#554	scope-violation	+59/-1583, 6 files	closed	Maybe splittable	Backend config plus dependency cleanup. The line count was dependency-heavy; config change and module cleanup could have split.
#556	scope-violation	+59/-1583, 6 files	closed	Maybe splittable	Same shape as #554. This looks like duplicate/retry friction rather than a threshold problem.
#622	scope-exception	+7/-4581, 49 files	merged	Legit-large / mechanical	Legacy demo removal. Large deletion is a valid exception class when the PR is deletion-only or near deletion-only.
#679	scope-exception	+15118/-2141, 23 files	merged	Legit-large / docs import	Dev Portal guide import is very large but docs-oriented. It should require an explicit docs-import reason and probably stay outside routine auto-ready.
#729	scope-violation	+881/-174, 7 files	closed	Maybe splittable	Dev Portal CSS, cache, docker compose, spec, and lockfile together. Split visual CSS from environment/lockfile updates.
#730	scope-violation	+879/-173, 4 files	closed	Maybe splittable	Similar CSS/system slice. Below hard limit but above soft warning; likely reviewability problem rather than false positive.
#733	scope-exception	+1047/-0, 3 files	merged	False positive / docs-only	Planning/spec docs exceeded the hard line limit but did not alter product runtime. Needs docs-only calibration, not a global threshold raise.
#819	scope-exception	+1157/-235, 5 files	merged	False positive / binary asset-heavy	Frontend redesign line count was inflated by a binary asset plus visual docs. The feature was coherent, but exception evidence should call out asset weight.
#825	scope-violation	+1097/-2, 4 files	closed	False positive / metadata-only	DB schema review docs plus workflow/test metadata. The hard line limit caught long analysis, not runtime risk.
#827	scope-violation	+1079/-0, 2 files	closed	False positive / docs-only	Discussion/planning docs only. Better handled by docs-specific exception guidance or issue comment when no repo artifact is required.
#844	scope-violation	+2823/-31, 14 files	closed	Splittable	Backend, frontend, schema, migration, tests, and docs in one PR. This is exactly what hard gates should block.
#845	scope-exception	+980/-10, 11 files	merged	Legit-large / migration	Backend-only raffle tiers API/model/migration with Atlas normalization. Under hard limit but still needs migration exception evidence.
#862	scope-violation, scope-exception	+0/-1704, 12 files	closed	Legit-large / deletion	Extension JWT backend removal was deletion-heavy and backend-only, but generated Swagger/test fallout should be listed in the exception reason.
#863	scope-exception	+0/-1739, 15 files	closed	Legit-large / deletion	Same deletion family as #862 with shared-types/contract checks included. Valid exception shape, but still needs explicit replacement/closure reason when not merged.

Recommendation

Keep the current global thresholds for normal product PRs:

• 600 diff lines should remain a warning.
• 1000 diff lines should remain a hard stop.
• 35 changed files should remain a hard stop.

Tighten scope-exception instead of raising the limits:

1. Require a short exception reason in the PR body or sticky comment category before the label is considered policy-compliant.
2. Recognize a small set of exception categories: mechanical-move, generated-output, large-deletion, migration-baseline, framework-deps, binary-asset, docs-import, and governance-policy.
3. Keep mixed product surfaces blocked unless the exception reason says why the surfaces are inseparable and names the follow-up split plan.
4. Treat docs-only or planning-only PRs above the hard line limit as a separate calibration problem. They may deserve docs-specific guidance, but not a broader product-code threshold raise.
5. Keep auto-ready away from exception PRs unless a human reviewer has already confirmed the exception reason and the risk class is not R4.

Follow-Up Cuts

This PR does not implement these changes. Possible follow-ups:

Follow-up	Scope	Validation
Exception reason taxonomy	Update docs/pr-scope-policy.md and PR template wording only.	git diff --check, PR metadata regression if template parsing changes.
Scope Police exception reason check	Add a lightweight check that warns or fails when scope-exception has no reason.	Workflow regression tests in .github/workflows/ci.test.ts.
Docs-only large-analysis guidance	Define when long docs belong in repo docs vs issue/discussion comments.	Docs diff check only.
Generated-output accounting	Teach sticky comments to identify generated Swagger / lockfile / migration output separately from source edits.	Workflow regression tests plus fixture PR metadata.