[amaybaum]

Draft — infrastructure only. Merging this PR and any dispatch of the workflow each need separate owner approval, after independent review.

Registers .github/workflows/k1-pers-1.yml as a dispatch-only stub for round K1-PERS-1 (provisional record: #914). GitHub registers a workflow_dispatch-only workflow only once the file is on the default branch. The precedent is 3c1aae8, which registered the PE-01 bounded-tests workflow on main the same way. The round's control plane later replaces this body as governed path .github/workflows/k1-pers-1.yml, op M.

What the stub does

• Trigger: on: workflow_dispatch only. It has no push, pull_request or schedule trigger and no inputs.

• Permissions: contents: read.

• Actions: pinned by commit, the same pins as pe01-bounded-tests.yml: actions/checkout v4.2.2 and actions/upload-artifact v4.6.2.

• Job: one job, smoke, a non-scientific smoke test with a 5-minute timeout. It:

  1. checks out the dispatched ref;
  2. prints the exact head SHA and fails if it differs from GITHUB_SHA;
  3. reports whether verification/programmes/oi-qm/reconstruction/round-k1-pers-1-carrier-persistence/ exists (absence is reported, not a failure);
  4. writes smoke.json (schema: k1-pers-1-smoke-v1, head_sha, smoke: true, scientific: false, round_directory_present), reads it back, checks it, and uploads it as the artifact k1-pers-1-smoke.

  The result is that marker plus the job's exit status, not log text (AGENTS.md §A.44).

What it does not do

• It imports and runs no probe, computation or round code. It makes no scientific claim, and it changes no manuscript, ledger, grade or roadmap.
• It does not start on push or pull request. It is not the round's F/E attestation; those use the repository's check workflow (§A.39). It is not the round's computational run.
• This PR opens no V3 round and designates nothing. The round's D is the head of main after this lands.

Trigger analysis

• No workflow at the base (07d94f2) has a push or pull_request path filter that matches .github/workflows/k1-pers-1.yml or the round directory. bg2-orientation-design.yml, bridge-compatibility-research.yml and hidden-sector-replay.yml filter only on their own record directories and their own workflow files.
• verify.yml has no path filters and runs on every pull_request, and on push to main. That is ordinary CI on this PR, not a governed launch.
• evidence-replay.yml and pe01-bounded-tests.yml are dispatch-only, so no push or PR starts them.
• No check in tools/ keeps a fixed list of workflow files. ci_gate_presence_test.py and coverage_check.py read only verify.yml, so no registry needs updating. The workflow lists in frozen audit snapshots (e.g. candidate-inputs.json) are records that no check reads.

Local checks (at the stub commit)

Check	Exit
tools/ci_gate_presence_test.py	0
tools/artifact_placement_check.py	0
papers/oi_lattice_code/architecture_check.py	0
tools/build_migration_manifest.py --check	0
PyYAML parse (triggers = {workflow_dispatch}, permissions = contents: read)	0

tools/legacy_records_check.py could not decide in the shallow local checkout and needs CI's full history.

The smoke steps were also simulated locally, with no dispatch, against the base commit, and the controls behaved as expected:

• a wrong GITHUB_SHA fails;
• a tree that contains the round directory reports round_directory_present: true.

Independent review: accept with nits (triggers, non-scientific content, correctness, repo conventions and messages all pass).

---

Generated by Claude Code

[amaybaum]

Serialized unit 2 record, part 1 of 2: read-only reassessment (verbatim; sha256 prefix 1274a24ebfd70c69). Its §7 draft closing comment was corrected after review: see part 2, findings 6–8, and the closing comment that follows.

---

Reassessment of draft PR #1030 (K1-PERS-1 workflow stub)

Read-only, 2026-11-02. PR #1030: branch infra/k1-pers-1-workflow-stub-20261101, head d62f40a9, base 07d94f2c,
draft, open. Compared against main at 7a4321c5. The base is an ancestor of main, 18 commits behind, and
.github/ is unchanged between them.

1. Original purpose

• The PR has one commit (d62f40a9) that adds one file, .github/workflows/k1-pers-1.yml (+62/−0). The PR has no
  issue comments, review comments or reviews; its body records an independent review of "accept with nits".
• It was written for the provisional K1-PERS-1 draft, OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 c.6562048958 §8 (draft lines 280–305): "a dispatch-only stub
  .github/workflows/k1-pers-1.yml is registered through a separate" PR on main first. The round would later
  modify the stub (governed op M), and stage R would be "the authorized dispatch at C1's commit". The precedent is
  3c1aae80 (PE-01 registration).
• The stub has a single non-scientific smoke job (k1-pers-1.yml:21–62). It checks the head SHA (:27–31), reports
  whether the round directory ROUND_DIR (:19) exists (:32–35), writes and verifies smoke.json (:36–57), and
  uploads it (:58–62). It runs no probe code (:10–11).

2. Trigger and effect

• Its only trigger is on: workflow_dispatch: (k1-pers-1.yml:12–13), with contents: read (:15–16). It has no
  push, pull_request or schedule trigger and no inputs.
• On main at 7a4321c5, no path-filtered workflow matches the file. bg2 (:4–11), bridge (:8–19) and
  hidden-sector (:11–22) filter only on their own directories and their own workflow files. evidence-replay and
  pe01 are dispatch-only.
• verify.yml:21–25 runs on every pull_request and on push to main. This is ordinary CI and it already ran
  on d62f40a9 (Lean kernel check, Mathlib bridge, Numerical probes: success).
• Leaving it open starts nothing. A further push to the branch would rerun only ordinary PR CI. The branch is
  not a special trigger branch (§A.43 item 4).
• Merging it would run verify.yml on main (ordinary). It would also register a dispatchable K1-PERS-1
  workflow on the default branch. That workflow would launch only by an explicit dispatch, which would itself need
  owner authorization.

3. Comparison with main (7a4321c5)

• main has no k1-pers-1.yml, and no tracked path or text mentions K1-PERS-1 or K1-RDY (git grep, ls-tree).
  The capability, a dispatchable K1-PERS-1 run, is therefore not provided on main.
• It is not needed, though. The stub's only consumer was stage R of K1-PERS-1, and the owner closed that arm "without
  freezing or executing the proposed round": "No computational inventory run is authorized" (OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 c.6563417740).
• F/E attestations never needed the stub. They use verify.yml's own workflow_dispatch (verify.yml:25; draft line
  304; PR body, "What it does not do").
• Verdict: the stub's purpose has lapsed (its round was closed unexecuted); the stub is neither satisfied by
  main nor redundant with it. It is not superseded by another workflow.

4. Dependants

• K1-RDY (PREREG as posted, K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6567522034; 320 lines): no occurrence of workflow, dispatch, .yml,
  K1-PERS or ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030.
  • Its execution clause (§5.5) is a written derivation plus, only if authorized, an exact enumeration that reports
    "through its exit status and a schema file (§A.44)". It names no workflow.
  • Its held-heads checkpoint (§9) lists PRs 1009, 1023, 1011, 1003, 918 and 1025, not ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030.
  • If K1-RDY execution is ever authorized and needs CI, it would need its own workflow or a verify.yml job. The stub
    hard-codes the K1-PERS-1 ROUND_DIR and schema k1-pers-1-smoke-v1 (k1-pers-1.yml:19, :44), so it would not
    serve K1-RDY as written.
• Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023, research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011, research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003, Introduction paper: reconcile sources and integrate scoped theorem recoveries #918: none of their bodies cites ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030 or k1-pers-1.yml. Their changed files
  (71, 5, 35, 3 and 285) include no .github/ or k1-pers path.
• Frozen records: none on main cites it (the grep in §3).
• Issue records citing it: OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 c.6562048958 (the provisional draft), c.6562546850, c.6563412734 ("Still held:
  … merges of … ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030"), c.6563412926 (the matrix row O-C6A: "stub ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030", "ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030 → O-C6A") and c.6563417740. All
  are coordination records; none is a frozen claim that depends on the stub.

5. Impact

• Serialized workflow (WIP limit 1): ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030 is not among the five holds and carries no active work. Leaving it
  open adds an apparently active draft to the queue. Closing it removes that and touches no hold.
• Independent review: merging would still need owner approval after review (PR body, first line). Closing makes
  no claim and adds nothing to main, so it needs no new review.
• The five holds (Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023, research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011, research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003, Introduction paper: reconcile sources and integrate scoped theorem recoveries #918) are unaffected either way. No head moves.

6. Unique content

• The branch holds 62 lines of a non-scientific smoke stub and nothing else. No surviving claim rests on it, so
  nothing must be preserved elsewhere (§A.42 item 5).
• Its design (pinned actions, a §A.44 marker) is already exemplified on main by pe01-bounded-tests.yml. Closing
  without merging keeps the branch, the commit and the PR anyway (§A.43 item 6).
• §A.43 item 6's "open an Issue carrying its open work" is satisfied by OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914: the PR has no open work of its own,
  since its parent arm is closed and its row is recorded at c.6563412926.

7. Recommended disposition: CLOSE UNMERGED

• Reason: its sole purpose was registering the dispatch target for K1-PERS-1's stage R, and the owner closed that arm
  unexecuted (c.6563417740). No surviving record or plan depends on it, and K1-RDY does not use it.
• Item 6's caveat ("a PR merely waiting for … a governed authorization is still an active PR") does not apply: the
  authorization it awaited can no longer arise for its round.
• RETAIN ON HOLD would have no release condition short of a new K1-PERS round, which would need its own reviewed
  workflow anyway (§3, §4).
• Proposed closing comment (post, then close as not planned without merging; do not delete the branch):

Closed without merging. Branch and commit are preserved.
This draft registered .github/workflows/k1-pers-1.yml (dispatch-only smoke stub, 62 lines, head d62f40a9) as
the dispatch target for round K1-PERS-1's computational stage (#914 c.6562048958 §8). The owner closed that arm as a
scoped negative paper result without freezing or executing it, with no computational inventory run authorized
(#914 c.6563417740), and asked for this PR to be reassessed (#914 c.6563412734).
Reassessment against main 7a4321c5:

• No record, plan or open PR depends on the stub. That includes K1-RDY (K1-RDY: paper-only readiness investigation under fixed selection rule #1031) and the held PRs Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023,
  research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011, research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003 and Introduction paper: reconcile sources and integrate scoped theorem recoveries #918.
• The F/E attestations use verify.yml's own workflow_dispatch.
• Nothing on the branch supports a surviving claim.

The workflow was never on the default branch and was never dispatched. Branch
infra/k1-pers-1-workflow-stub-20261101 and commit d62f40a9046a1bed22dde03a081fd85abf508c61 stay as provenance.
Any future computational round registers its own reviewed workflow. Tracking for row O-C6A stays on #914.

---

Generated by Claude Code

[amaybaum]

Serialized unit 2 record, part 2 of 2: independent review with confirmation (verbatim; sha256 prefix b55673495e9c913d).

---

Independent review: reassessment of draft PR #1030

Read-only, 2026-11-02. Checked against PR head d62f40a9, base 07d94f2c and main 7a4321c5. I used the
GitHub REST API (GET only) and git fetch / show / diff / grep with gc and maintenance disabled.

Verdict: CONFIRM-WITH-NITS. CLOSE UNMERGED is the right disposition. Before §7's comment is posted, it needs
the two fixes in findings 6 and 7.

1. Diff. API and git agree: one commit, one file added, .github/workflows/k1-pers-1.yml, +62/−0. The only
   trigger is on: workflow_dispatch: (lines 12–13), with no inputs. Permissions are contents: read (15–16). The
   PR has 0 issue comments, 0 reviews and 0 review comments. The base is an ancestor of main, 18 commits back, and
   .github/ is unchanged between them.
2. Triggers. Of main's six workflows, bg2 and bridge fire on push to any branch and on pull_request, but
   only for their own paths. hidden-sector is filtered to main and its own paths. evidence-replay and pe01 are
   dispatch-only. verify.yml (lines 21–25) runs on every pull_request. The branch has exactly one run: verify,
   pull_request, success. The repository's workflow registry has no K1-PERS-1 entry, so it was never registered
   or dispatched. Merging would run ordinary verify CI on main and register a dispatch-only workflow. No tool on
   main lists workflow files: the three readers read only verify.yml. Merging would also put stale text on main:
   "the round's control plane replaces this body".
3. Main. A case-insensitive git grep of main for k1-pers / k1_pers / K1-RDY finds nothing (exit 1).
   The stub's ROUND_DIR does not exist on main.
4. OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 citations. c.6563417740 (owner decision) matches: the arm is closed "without freezing or executing",
   "No computational inventory run is authorized", and "ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030 stays unmerged … will be reassessed".
   c.6563412734 says only "Still held: … merges of #924 correction (stage 1): Barandes citation labels (M3) and Explainer class statement (M4) #1026, #1004 correction: remove the θ = 0 claim from ch02 and align Structure Q12 with SM §8.7 #1027, Research audit: observer-relative finiteness and finite-carrier proof boundaries #1017 and ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030". It does not contain the
   reassessment request. §4's quotation of it is accurate. The §8 draft lines 280–305 of c.6562048958 and the O-C6A
   row of c.6563412926 are accurately cited.
5. Dependants. K1-RDY appears both as the local PREREG (313 lines) and as posted (K1-RDY: paper-only readiness investigation under fixed selection rule #1031 c.6567522034, 320
   lines). Neither mentions a workflow, dispatch, .yml, K1-PERS or ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030. §5.5 requires only exit status plus a
   schema file. The held-heads row is in §9, line 263. The bodies and paginated file lists of Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023, research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011,
   research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003 and Introduction paper: reconcile sources and integrate scoped theorem recoveries #918 contain no .github/ or k1-pers path. A future K1-RDY execution is not harmed by closing:
   the stub hard-codes K1-PERS-1's ROUND_DIR and schema, any run needs a new reviewed workflow anyway, and the
   branch is kept.
6. Misattribution in §7 (fix). "asked for this PR to be reassessed (OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 c.6563412734)" should cite
   c.6563417740. Alternatively, cite c.6563412734 only for holding the merge.
7. Footer missing in §7 (fix). The proposed comment does not end with --- /
   _Generated by [Claude Code](https://claude.ai/code)_. Append it.
8. Nit. "Close as not planned" is a state reason for Issues. A PR is simply closed. Otherwise the comment
   is accurate: it states "without merging", says the branch and SHA are preserved, and makes no claim beyond what
   was checked here.
9. §A.43 item 6. OI ↔ QM equivalence: proof gate (dependency index; carries #885) #914 already carries the decision and the O-C6A row and cites ci: register the K1-PERS-1 workflow as a dispatch-only stub (infrastructure only) #1030, and the closing comment
   gives the branch and full SHA. No new Issue is needed. The "awaiting authorization" caveat does not apply,
   because the round the stub served is closed.
10. Bias. RETAIN ON HOLD has no release condition: K1-PERS-1 is closed and K1-RDY cannot use the stub. The
    owner's wording ("will be reassessed") is neutral, and the evidence supports closing on its own merits. I find no
    sign the recommendation was driven by the owner's preference.

Confirmation. ACCEPT. The corrected closing comment cites c.6563417740 for the reassessment, ends with the
required footer and closes plainly without merging; its claims match findings 1–5, with no residual dependency.

---

Generated by Claude Code

[amaybaum]

Closed without merging. The branch and commit are preserved.

This draft adds one file, .github/workflows/k1-pers-1.yml (62 lines, head d62f40a9). It is a dispatch-only smoke stub with contents: read, and it was registered as the dispatch target for the computational stage of round K1-PERS-1 (#914 c.6562048958 §8).

That arm was closed as a scoped negative paper result, without being frozen or executed and with no computational inventory run authorized. The same closure record (#914 c.6563417740) kept this PR unmerged pending reassessment.

The reassessment was made against main 7a4321c5 and independently reviewed. It found:

• No record, plan or open PR depends on the stub. This includes K1-RDY (K1-RDY: paper-only readiness investigation under fixed selection rule #1031) and the held PRs Research (WIP): SUBSTATE-1 — internal sub-states and visible recurrence #1009, Research (draft): MR-1 — measurement as physical record formation, preregistration only #1023, research: SUBSTATE-2 recursive inverse-limit substructure and depth-feedback controls #1011, research: OI ↔ constructive QCD bridge and gauge-action non-uniqueness #1003 and Introduction paper: reconcile sources and integrate scoped theorem recoveries #918.
• F/E attestations use verify.yml's own workflow_dispatch.
• Nothing on the branch supports a surviving claim.

The workflow was never on the default branch and was never dispatched. Branch infra/k1-pers-1-workflow-stub-20261101 and commit d62f40a9046a1bed22dde03a081fd85abf508c61 remain as provenance. Any future computational round would register its own reviewed workflow. Tracking for row O-C6A stays on #914.

---

Generated by Claude Code