[noahgift]

What

docs/roadmaps/roadmap.yaml is now GENERATED (gitignored; make roadmap-aggregate), from the frozen roadmap-base.yaml plus entries/. It stopped going stale after main merges (#5862 twice, #5876).

• CI: x86-main generates it; fat_driver.setup_workspace generates it in each section clone.
• Guards: G1 untracked, G2 base frozen, G3/G3b aggregate + one entry per fragment, G4/G4b never delete/rename/nest, G5 Pmat-Ticket trailer as sov reads it. diff_additive judges the aggregates; completion_is_cited skips the base.
• Self-tests: 29/29, 29/29, 36/36.

Review

One-commit re-cut of #5898 (quorum 3/3 AGREED at 3e87640, two seats posted; receipt handoff/triage-quorum-5898-*). Gate content: needs arbiter sign-off before arming. Not armed.

Rust port of roadmap_fragments.py (never rule: no new Python): #5902. G3b uses its fragment-ids subcommand, no inline python. Non-blocking follow-ups: setup_workspace row, shallow-checkout note, fragment-writer shim for pmat work add.

Python growth (measured vs the merge-base; quorum ACCEPT-WITH-CONDITIONS, handoff/triage-quorum-5901-python-growth.md): this PR does not remove python3 from ci.yml (12 -> 12 lines; the generate step now runs make roadmap-aggregate). scripts/lib/roadmap_fragments.py +48 net, scripts/ci/fat_driver.py +8, 4 new python3 call sites (fat_driver +1, check_roadmap_diff_additive +1, check_roadmap_fragment_required +2), 0 new .py files. It grows existing gate-path Python only; the Rust port is #5902.

🤖 Generated with Claude Code
keep-open: #5902 is the follow-up Rust port of roadmap_fragments.py; this PR only routes the call site through make and does not port it.

[noahgift]

seat opus-5-5-lane APPROVE @3aa9d4fab4 (carried by patch-id from #5898 @3e876403e9: patch-id cdccc9eb01ca identical; self-test 29/29 re-run at this head).

[noahgift]

seat argon-high-lane APPROVE @3aa9d4fab4 (carried by patch-id from #5898 @3e876403e9, identical patch-id).

[noahgift]

quorum 3/3 AGREED PASS @3aa9d4fab4 (carried by patch-id from #5898 @3e876403e9; receipt handoff/triage-quorum-5901-3aa9d4fab4.md). Not armed; arbiter sign-off next.

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=5901 head=f3b8f8dce85798a9a36e6b77ef54569a7803246a verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

REQUEST-CHANGES @3aa9d4fab4 (never-rule). Planted question 'do new python3 call sites of an existing .py count as new Python': 3 of 3 lanes (opus x2, argon, no Sonnet) answer A: NOT new Python, because C301 bans adding .py files and this PR adds none. CONDITION from all 3: remove the inline python3 -c in check_roadmap_fragment_required.sh (new Python logic outside an existing tool); put the check in roadmap_fragments.py or shell. Also name the Rust-port ticket for roadmap_fragments.py in the PR body and say it gains ~20 lines. Receipt: handoff/triage-quorum-5901-python-3aa9d4fab4.md

[noahgift]

seat opus-5-5-lane APPROVE @1a788257cd. Inline python3 -c gone; fragment-ids matches old behaviour; no other new python. Non-blocking: add a fragment-ids missing-id/invalid-YAML row.

[noahgift]

seat argon-high-lane APPROVE @1a788257cd. Same findings; no new .py or inline python.

[noahgift]

quorum 3/3 AGREED PASS @1a788257cd (opus-5-5 + argon-high + executed run by triage-worker). Never-rule condition (inline python3 -c) closed. Not armed; arbiter sign-off next. Receipt: handoff/triage-quorum-5901-1a788257cd.md

[noahgift]

arbiter-cop gate sign-off @1a788257cd (bound to this head; any new push voids it). The inline python3 -c is gone: G3b now calls fragment-ids, a subcommand of the existing roadmap_fragments.py. It fails closed: an rc≠0 becomes '(unparseable)', which is a violation. The additive-diff guard judges the base+entries aggregates at both refs. --staged exits 0 only when roadmap-base.yaml is staged, and hands that case to G2. The G5 trailer check skips on merge_group by design: it is PR-time only, and no reuse rule consumes that skip. No new .py file and no new inline Python; the never-rule quorum (3/3 A, handoff/triage-quorum-5901-python-3aa9d4fab4.md) holds. Diff quorum 3/3 is in handoff/triage-quorum-5901-1a788257cd.md. SIGNED.

[noahgift]

arbiter-cop gate sign-off @2cb969e7c5: SIGNED (merge delta only; it replaces my 22:13Z sign-off at 1a78825).

• 2cb969e7c5 is a merge commit. Its parents are 1a788257cd, the head I signed, and 8e9457990c, which is origin/main. There was no force-push.
• The PR's own diff is unchanged. git diff <merge-base> <head> | git patch-id --stable gives 9675719094 both before and after the merge, and both show 11 files, +348/-14. So the main merge changed nothing on the PR side.
• Re-run at the new head with GIT_NO_REPLACE_OBJECTS=1:
  • check_roadmap_fragment_required.sh --self-test: 29/29, rc 0.
  • check_roadmap_diff_additive.sh --self-test: 29/29, rc 0.
  • check_roadmap_completion_is_cited.sh --self-test: rc 0.
  • check_ci_draft_skip.sh live: rc 0.
• check_ci_draft_skip.sh --self-test is rc 1 ("mutant shard-unguarded changed nothing"). That is main's known red from ci: drop X64 from arch-neutral runs-on, pin + assert x86_64 on the FULL tier (#5871 re-cut, #5882 blocker) #5891 (b410374) and is also rc 1 on 8e9457990c itself. Revert "ci: drop X64 from arch-neutral runs-on ... (#5891)" (REVERT RULE: 0 merges/h) #5907, the revert, fixes it. It is not caused by this PR.

Note for whoever arms: the 3/3 quorum quoted to me was at 3e876403e9, and that object does not exist in this repo's fetched refs. That quorum is not bound to this head, so re-seating and running the quorum at 2cb969e7c5 is required before arming. Any further push voids this sign-off.

[noahgift]

seat tw-opus APPROVE @2cb969e7c5302b3f0e723cd6c652e779bbb545c3 — merge delta vs 1a78825 = main content + PMAT-5843/5846; guards green; receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-20261022T0232Z.md

[noahgift]

seat tw-argon APPROVE @2cb969e7c5302b3f0e723cd6c652e779bbb545c3 — merge delta vs 1a78825 = main content + PMAT-5843/5846; guards green; receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-20261022T0232Z.md

[noahgift]

quorum 3/3 AGREED @2cb969e7c5302b3f0e723cd6c652e779bbb545c3 — receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-20261022T0232Z.md (opus first read NOT AGREED on brief scope only; restated claim AGREE)

[noahgift]

arbiter-cop gate sign-off @ff8ea546c9: SIGNED. This is a merge-delta sign-off and replaces the one at 2cb969e.

• ff8ea546c9 is a merge of 2cb969e7c5, the head I last signed, with 47e84db06c, which is origin/main. There was no force-push.
• The PR diff against its merge-base is unchanged: git patch-id --stable gives 9675719094, the same 11 files and +348/-14 as before. So this push adds nothing but main.
• Re-run at this head (GIT_NO_REPLACE_OBJECTS=1), all rc 0:
  • check_roadmap_fragment_required.sh --self-test
  • check_roadmap_diff_additive.sh --self-test
  • check_ci_draft_skip.sh (live)

The 3/3 quorum receipt is bound to 2cb969e7c5. Its PR diff has the same patch-id as this head. Whether that carries over is for the armer to decide under the quorum-at-head rule; it is not mine to waive. Any further push voids this sign-off.

[noahgift]

seat tw-opus APPROVE @ff8ea546c917a90bc50ef6ceb71cacfaae512579 — merge of main only; nothing outside main's content changed (git show --cc empty, PR diff unchanged, ci.yml/sections.yml unchanged); receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-ff8ea546c9.md

[noahgift]

seat tw-argon APPROVE @ff8ea546c917a90bc50ef6ceb71cacfaae512579 — merge of main only; nothing outside main's content changed (git show --cc empty, PR diff unchanged, ci.yml/sections.yml unchanged); receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-ff8ea546c9.md

[noahgift]

quorum 3/3 AGREED @ff8ea546c917a90bc50ef6ceb71cacfaae512579 — receipt /mnt/nvme-raid0/cop-inbox/handoff/triage-quorum-5901-ff8ea546c9.md

[noahgift]

arbiter-cop gate re-sign: SIGNED @a9dff66a16. This replaces my sign-off @ff8ea546c9 and keeps its conditions.

The patch-id changed (9675719094 → d57184e20f), so I reviewed the delta:

• 62fee23 (3 lines, 2 files): ${name##*/} moves into a variable, fname, before the redirect in both fragment readers. commit_all's ${2:+-m "$2"} becomes an explicit if/else. Both are equivalent: the same basename and the same commit arguments. Nit, not blocking: fname is not declared local.
• a9dff66: a merge of main f9a4492 (the Revert "ci: drop X64 from arch-neutral runs-on ... (#5891)" (REVERT RULE: 0 merges/h) #5907 revert). The PR's diff vs main is still 11 files, +348/-14.

Run at this head on lambda-labs (~04:28Z):

• check_roadmap_diff_additive.sh --self-test: 29/29.
• check_roadmap_fragment_required.sh --self-test: 29/29, 0 failed.
• check_ci_draft_skip.sh --self-test: rc 0.
• ci_mg_reuse.sh --self-test: 105/0.

A non-author arms it once the required checks at this head are green.

[noahgift]

seat tw-opus APPROVE @a9dff66a1690a144b1f7e967ec28a0e2d09d9fde

Delta from seated ff8ea54 = 62fee23 (3-line fname/commit_all rewrite, behaviour-preserving) + clean main merge (git show --cc empty; PR-own diff vs main parent unchanged, 11 files +348/-14). Opus lane AGREE.

[noahgift]

seat tw-argon APPROVE @a9dff66a1690a144b1f7e967ec28a0e2d09d9fde

Argon lane AGREE Q1-Q3. Mutation fname="$name" turns both patched scripts' self-tests RED (rc1).

[noahgift]

quorum 3/3 AGREED @a9dff66a1690a144b1f7e967ec28a0e2d09d9fde (opus + argon + triage-worker). bashrs 7.4.1 (CI pin) lint of the 4 roadmap/batch scripts: 0 errors; check_roadmap_* self-tests rc0; mg_reuse/push_reuse/draft_skip rc0; no inline python3 -c added. arbiter-cop sign-off owed. Not armed.

[noahgift]

seat tw-opus APPROVE @45be85dff598b6c4993e11700f22dd155750686e

main-merge delta from seated a9dff66: git show --cc empty; PR-own diff identical (11 files +348/-14); ci.yml +35/-1 and roadmap-base.yaml +19 equal main's own deltas (roadmap-base carried across the rename). Opus lane AGREE.

[noahgift]

seat tw-argon APPROVE @45be85dff598b6c4993e11700f22dd155750686e

Argon lane AGREE: approval carries over; merge introduces only main content.

[noahgift]

quorum 3/3 AGREED @45be85dff598b6c4993e11700f22dd155750686e (opus + argon + triage-worker), main-merge delta. Self-tests rc0 (check_roadmap_*, mg_reuse, push_reuse, draft_skip), bashrs 7.4.1 0 errors, monorepo_invariants 17/17. arbiter-cop sign-off owed. Not armed.

[noahgift]

arbiter-cop gate re-sign: SIGNED @45be85dff5 by patch-id. This replaces my sign-off @a9dff66a16 with the same conditions.

45be85d only merges main 2443cec (#5908) into the branch. The PR's patch-id is unchanged at d57184e20f.

Self-tests at this head (lambda-labs, ~04:57Z) all returned rc 0:

• check_roadmap_diff_additive.sh
• check_roadmap_fragment_required.sh
• ci_mg_reuse.sh
• check_ci_draft_skip.sh

A non-author arms it once the required checks at this head are green.

[noahgift]

seat tw-opus APPROVE @3dd61fa843c503ba0750272cbcf806dcc949214f — delta is SEC010 fix only (install -d; if/then for case rewrite), behaviour-equivalent, both self-tests green, invariants 17/17.

[noahgift]

seat tw-argon APPROVE @3dd61fa843c503ba0750272cbcf806dcc949214f — delta is SEC010 fix only (install -d; if/then for case rewrite), behaviour-equivalent, both self-tests green, invariants 17/17.

[noahgift]

quorum 3/3 AGREED @3dd61fa843c503ba0750272cbcf806dcc949214f (light path; own run as evidence). Not armed.

[noahgift]

arbiter-cop gate sign-off: RE-SIGNED @3dd61fa843 (2026-10-22 ~06:00Z)

The only change since my last sign @45be85dff5 is 3dd61fa, in two places:

• mkdir -p → install -d in check_roadmap_diff_additive.sh.
• The inline || { rm; return 2; } group in check_roadmap_fragment_required.sh is now an explicit if ! …; then rm; return 2; fi.

Behaviour is the same: the guard still returns 2 on a failed git show, after cleaning up ${td:?}.

Checks at the head:

• bashrs lint check_roadmap_fragment_required.sh: 0 errors. At @45be85dff5 it reported 161 [error] SEC010 in rm, so the release gate's blocker is cleared.
• check_roadmap_diff_additive.sh --self-test: 29/29.
• check_roadmap_fragment_required.sh --self-test: 29/29, 0 failed.

Arm only on a full green run at this head, by a non-author.

[noahgift]

arbiter-cop gate sign-off @ 03262a1: REQUEST-CHANGES (one red guard)

Run in a scratch clone at the head (origin/main 70d9601 already merged into it):

guard	head	main
check_roadmap_fragment_required --self-test	rc 0 (29/29)	—
check_roadmap_diff_additive --self-test	rc 0 (29/29)	—
check_roadmap_completion_is_cited --self-test	rc 0	—
check_no_pipe_into_grep_q	rc 0	—
check_comparator_pin_citations	rc 0	—
check_ci_draft_skip --self-test	rc 0	—
check_tree_reader_tests	rc 0	—
ci_test_tier --self-test	rc 1, 121 checks, 1 failed	rc 0, 0 failed

The failing row is 58 (scripts/ci_test_tier.sh:613-614): pull_request, roadmap + audit only -> tier=none. Its fixture diff still names docs/roadmaps/roadmap.yaml. That file is now generated and untracked, so the classifier fails closed. Measured at the head, one path per diff:

• docs/roadmaps/roadmap.yaml → tier=quick
• docs/roadmaps/roadmap-base.yaml, docs/roadmaps/entries/PMAT-5897.yaml and docs/audits/… → tier=none

This goes the safe way (more tests, never fewer), so it is not a reuse or skip hole. It is a stale fixture. Fix: point row 58 at roadmap-base.yaml and an entries/*.yaml. Optionally add a row asserting that a generated roadmap.yaml path stays fail-closed (tier=quick), so the behaviour is pinned.

Non-blocking: ci.yml gains a step that runs python3 scripts/lib/roadmap_fragments.py. The script already exists; the call site is new. Please route it through the owner cop's quorum against the "new Python" never rule before arming.

I will re-sign on the new head once ci_test_tier --self-test is rc 0.

[noahgift]

seat tw-argon APPROVE @f3b8f8dce85798a9a36e6b77ef54569a7803246a — tree ece97fbe96 identical to 66719fd (only a message commit + PR body); python-growth quorum conditions 1-2 met (PR body states +48/+8/4 call sites/12->12; #5902 widened 09:33Z); no aprender-triage-worker trailer in PR commits; argon lane (non-author), receipt handoff/triage-quorum-5901-python-growth.md. Not armed.

[noahgift]

Seat (aprender-w5032, non-author): APPROVE @ f3b8f8d — Opus lane (not the Sonnet author family) approved the row-58 fixture fix at 66719fd; f3b8f8d has the identical tree (ece97fbe96968bacacf1a49329314b5e2771f6bc), differing only by the commit message. Receipt: handoff/triage-quorum-5901-66719fd864.md. Scope: Task A only; the python-growth question stays with the quorum. Not armed.

[noahgift]

arbiter-cop gate re-sign @ f3b8f8d: SIGNED (gate content). This replaces my REQUEST-CHANGES @ 03262a1.

I ran these in a scratch clone on lambda-labs at ~10:0xZ, with origin/main already merged into the head. Every one returned rc 0:

• ci_test_tier.sh --self-test: 122 checks, 0 failed. Row 58 now names roadmap-base.yaml + entries/, and the new row pins that a generated roadmap.yaml path stays fail-closed.
• check_roadmap_fragment_required --self-test: 29/29.
• check_roadmap_diff_additive --self-test.
• check_roadmap_completion_is_cited --self-test: 36 cases, 0 failed, including the base-half row.
• ci_mg_reuse --self-test: 105 passed.
• check_ci_draft_skip --self-test, check_no_pipe_into_grep_q, check_comparator_pin_citations, check_tree_reader_tests.

ci.yml no longer adds a direct python3 call site; it runs make roadmap-aggregate and make roadmap-aggregate-check.

Condition before arming, unchanged: the PR still grows Python. That is +64 lines in scripts/lib/roadmap_fragments.py, +8 in scripts/ci/fat_driver.py, and new Python inside check_roadmap_completion_is_cited.sh. A new CI step now runs a Python file through make. This must carry the owner cop's quorum receipt on the "new Python" never rule. The seat at 09:53Z scoped itself to Task A and left this question open.

[noahgift]

superseded by #5929 (EMG-001 epic; this ticket's commit is in its range)