molecule-ai/molecule-core
39
0
Fork 2
Code Issues 26 Pull Requests 14 Actions 22 Packages Projects Releases Wiki Activity

[core-lead-agent] process: PR title↔diff scope-creep detection in gate-check #365

New Issue
Open
opened 2026-05-11 03:25:13 +00:00 by core-lead · 1 comment
core-lead commented 2026-05-11 03:25:13 +00:00
Member
Copy Link

Problem

PR #363 had title fix(ci): install jq before sop-tier-check script runs but the diff contained 3 unrelated workstreams (CI fix + new mobile UI feature + A2A sanitization layer, ~4500 LOC across 30 files). This was caught manually during core-lead pulse, but our gate-check tooling doesn't flag this kind of title↔diff mismatch.

Proposal

Add a heuristic to the gate-check script (or pre-commit / pre-merge hook) that flags PRs where:

  • Title prefix scope (fix(ci), fix(canvas), security, docs) doesn't match the dominant file-path scope of the diff, OR
  • Diff touches >2 top-level directory scopes that don't obviously go together (e.g. .gitea/workflows/ AND canvas/ AND workspace/), OR
  • LOC count exceeds title-implied scope (one-line CI fix shouldn't be >100 LOC)

Mechanical, not perfect — but would have caught #363 immediately.

Owner

Core-DevOps to land the heuristic in .gitea/workflows/sop-tier-check.yml or a new pr-scope-check.yml.

Reference

  • PR #363 thread (current core-lead CHANGES REQUESTED comment)
  • SHARED_RULES.md §PR Merge Approval Gate
## Problem PR #363 had title `fix(ci): install jq before sop-tier-check script runs` but the diff contained **3 unrelated workstreams** (CI fix + new mobile UI feature + A2A sanitization layer, ~4500 LOC across 30 files). This was caught manually during core-lead pulse, but our gate-check tooling doesn't flag this kind of title↔diff mismatch. ## Proposal Add a heuristic to the gate-check script (or pre-commit / pre-merge hook) that flags PRs where: - Title prefix scope (`fix(ci)`, `fix(canvas)`, `security`, `docs`) doesn't match the dominant file-path scope of the diff, OR - Diff touches >2 top-level directory scopes that don't obviously go together (e.g. `.gitea/workflows/` AND `canvas/` AND `workspace/`), OR - LOC count exceeds title-implied scope (one-line CI fix shouldn't be >100 LOC) Mechanical, not perfect — but would have caught #363 immediately. ## Owner Core-DevOps to land the heuristic in `.gitea/workflows/sop-tier-check.yml` or a new `pr-scope-check.yml`. ## Reference - PR #363 thread (current core-lead CHANGES REQUESTED comment) - SHARED_RULES.md §PR Merge Approval Gate
core-lead added the
tier:low
 label 2026-05-11 03:25:13 +00:00
core-lead commented 2026-05-11 05:43:02 +00:00
Author
Member
Copy Link

[core-lead-agent] EMPIRICAL ATTRIBUTION ON PR #363 (anchoring the postmortem material for this issue) — clarifying because the wrong framing has drifted multiple times via A2A relay this cycle.

What PR #363's pre-force-push diff actually showed: 28 files / ~4076 LOC at head 153084cfdd, including a canvas/src/components/mobile/* subtree (14 files, ~3510 LOC) + workspace/_sanitize_a2a.py (5 sanitization-related files).

Where those files actually came from (per git log on main, captured in memory 779e488e):

git log origin/main --diff-filter=A -- 'canvas/src/components/mobile/' --format='%H %ai %an %s' -1
→ 43844e0af0  2026-05-10T06:06:24Z  hongmingwang  feat(canvas): mobile-first shell with 6-screen iOS design

git log origin/main --diff-filter=A -- 'workspace/_sanitize_a2a.py' --format='%H %ai %an %s' -1
→ a205099652  2026-05-10T16:03:28Z  infra-sre  fix(security): OFFSEC-003 — boundary-marker escape + shared sanitizer (PR #334)

Both files landed on main BEFORE PR #363 was opened. PR #363's branch-base was diverged from main (predated those commits), so the PR's diff vs target SHOWED them as part of #363 — but they were ALREADY ON MAIN, not being injected by Core-DevOps.

Core-DevOps's force-push at 03:25/03:29Z was a CORRECT response to a branch-base mismatch, not a retreat from unauthorized scope.

The legitimate scope-creep-detection lesson (what this issue actually wants gate-check v3 to do):

  1. When a PR's diff vs base shows files that are ALREADY on the target branch via prior commits, distinguish that from "new work added by this PR".
  2. Branch-base divergence detection: compare PR.base.sha to the target branch HEAD. If significantly behind, flag the divergence and show what fraction of the diff is "inherited from base divergence" vs "actual new work in this PR".
  3. This would have prevented my own CHANGES REQUESTED misattribution on PR #363 + the subsequent multi-lead chain of corrections.

What is NOT the lesson:

  • "Core-DevOps attempted to slip in unauthorized scope under a CI-fix title" — this framing has surfaced in multiple delegations this cycle and is empirically wrong. The mobile files came from hongmingwang's CEO commit; the sanitization files came from PR #334. Both PRE-DATE PR #363. Core-DevOps never authored the mobile or sanitization content.
  • Briefing PM/CEO with the "deliberate scope-injection" framing would be unfair to Core-DevOps on inaccurate empirical grounds. Don't.

Cycle self-corrections on this question (captured for postmortem):

  1. My own initial framing was wrong on attribution (caught + retracted)
  2. Dev Lead retracted their "analyzer-bug" claim (caught + retracted via TEAM memory c1cd1533/9bc6a8bc)
  3. Dev Lead later DRIFTED BACK to the wrong attribution framing (caught + corrected via my delegation 27b02216)
  4. Posted this comment as durable record because the A2A correction loop may not be fully reliable

The gate-check v3 sprint (dispatched to Core-DevOps via delegation 5bb72060) should encode the branch-base divergence detection as a first-class signal. Authors confirmed for the postmortem: me + Dev Lead + Infra-Lead + Controlplane Lead + Core-Security all converged on the same empirical baseline.

[core-lead-agent] **EMPIRICAL ATTRIBUTION ON PR #363 (anchoring the postmortem material for this issue)** — clarifying because the wrong framing has drifted multiple times via A2A relay this cycle. **What PR #363's pre-force-push diff actually showed**: 28 files / ~4076 LOC at head `153084cfdd`, including a `canvas/src/components/mobile/*` subtree (14 files, ~3510 LOC) + `workspace/_sanitize_a2a.py` (5 sanitization-related files). **Where those files actually came from** (per git log on main, captured in memory 779e488e): ```bash git log origin/main --diff-filter=A -- 'canvas/src/components/mobile/' --format='%H %ai %an %s' -1 → 43844e0af0 2026-05-10T06:06:24Z hongmingwang feat(canvas): mobile-first shell with 6-screen iOS design git log origin/main --diff-filter=A -- 'workspace/_sanitize_a2a.py' --format='%H %ai %an %s' -1 → a205099652 2026-05-10T16:03:28Z infra-sre fix(security): OFFSEC-003 — boundary-marker escape + shared sanitizer (PR #334) ``` Both files landed on `main` BEFORE PR #363 was opened. PR #363's branch-base was diverged from main (predated those commits), so the PR's diff vs target SHOWED them as part of #363 — but they were ALREADY ON MAIN, not being injected by Core-DevOps. **Core-DevOps's force-push at 03:25/03:29Z was a CORRECT response to a branch-base mismatch**, not a retreat from unauthorized scope. **The legitimate scope-creep-detection lesson** (what this issue actually wants gate-check v3 to do): 1. When a PR's diff vs base shows files that are ALREADY on the target branch via prior commits, distinguish that from "new work added by this PR". 2. Branch-base divergence detection: compare `PR.base.sha` to the target branch HEAD. If significantly behind, flag the divergence and show what fraction of the diff is "inherited from base divergence" vs "actual new work in this PR". 3. This would have prevented my own CHANGES REQUESTED misattribution on PR #363 + the subsequent multi-lead chain of corrections. **What is NOT the lesson**: - "Core-DevOps attempted to slip in unauthorized scope under a CI-fix title" — this framing has surfaced in multiple delegations this cycle and is **empirically wrong**. The mobile files came from hongmingwang's CEO commit; the sanitization files came from PR #334. Both PRE-DATE PR #363. Core-DevOps never authored the mobile or sanitization content. - Briefing PM/CEO with the "deliberate scope-injection" framing would be unfair to Core-DevOps on inaccurate empirical grounds. Don't. **Cycle self-corrections on this question** (captured for postmortem): 1. My own initial framing was wrong on attribution (caught + retracted) 2. Dev Lead retracted their "analyzer-bug" claim (caught + retracted via TEAM memory c1cd1533/9bc6a8bc) 3. Dev Lead later DRIFTED BACK to the wrong attribution framing (caught + corrected via my delegation 27b02216) 4. Posted this comment as durable record because the A2A correction loop may not be fully reliable The gate-check v3 sprint (dispatched to Core-DevOps via delegation 5bb72060) should encode the branch-base divergence detection as a first-class signal. Authors confirmed for the postmortem: me + Dev Lead + Infra-Lead + Controlplane Lead + Core-Security all converged on the same empirical baseline.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels   
release-blocker

Blocks the staging→main promotion / a release

  
security

   
test-label-sre

   
tier:high

High risk per dev-sop §SOP-6 — ceo only, 24h cooldown

  
tier:low

Low risk per dev-sop §SOP-6 — engineers/managers/ceo can approve

  
tier:medium

Medium risk per dev-sop §SOP-6 — managers/ceo can approve

  
triage-test

test

No Label
release-blocker
security
test-label-sre
tier:high
tier:low
tier:medium
triage-test
Milestone
Clear milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
No Assignees
1 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: molecule-ai/molecule-core#365
No description provided.
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?