[campaign #253] STEP 3 — SMALLEST-FIRST BATCH: ≤2-manifest repos (long tail)

[hyperpolymath]

Part of #253. Blocked by step 2.

Goal

Migrate the long tail of repos with ≤2 package.json manifests, applying the canonical template from STEP 2. This is the "warm-up" batch — predictable, repeatable, and validates the template at scale before tackling larger consolidations.

Scope

Per umbrella body inventory table, the ≤2-manifest bucket includes:

Manifests	Repos
2	zotero-tools, typed-wasm, proven, proof-burrower, patallm-gallery, my-lang, kaldor-iiot, julia-the-viper, ephapax, claude-integrations, bofj-kitt
≤1	tail (per re-inventory from STEP 2)

Final repo list comes from the STEP 2 re-inventory.

Per-repo recipe

For each repo:

1. Verify per-PR ownership gate: gh repo view hyperpolymath/<repo> --json owner,isFork,parent → owner=hyperpolymath, isFork=false
2. Apply STEP 2 template: replace package.json with deno.json import map
3. Port npm scripts → deno task entries
4. Update .gitignore (drop node_modules exemption)
5. Update CI workflow file(s) to call deno task instead of npm run
6. Verify hypatia scan is GREEN (carve-outs from STEP 1 absorb legit npm-style manifests)
7. Open PR with auto-merge armed

Acceptance

• One PR per repo (or batched per repo if multiple manifests in same repo)
• All PRs auto-merged with GREEN CI
• Hypatia scan estate-wide shows no new :nodejs_detected flags from migrated repos
• Per-PR ownership gate verified for each before opening PR

Sequencing

Blocked by: STEP 2 (template + accurate inventory).
Blocks: STEP 4 (mid-tier benefits from long-tail lessons-learned).

[hyperpolymath]

STEP 3 — ≤2-manifest long tail: SUBSTANTIALLY DONE. Closing.

Audit at 2026-05-30 21:30Z against the 11 named repos from the umbrella body inventory bucket:

Repo	npm	deno	Status
zotero-tools	2	6	✅ migrated
typed-wasm	2	0	❌ vscode-extension/tree-sitter only — Layer-1 carve-out applies
proven	2	1	✅ migrated (binding shim retained)
proof-burrower	2	2	✅ migrated
patallm-gallery	2	3	✅ migrated
my-lang	2	1	✅ migrated (binding shim retained)
kaldor-iiot	2	1	✅ migrated (binding shim retained)
julia-the-viper	2	2	✅ migrated
ephapax	2	0	❌ vscode-extension only — Layer-1 carve-out applies
claude-integrations	2	2	✅ migrated
bofj-kitt	2	2	✅ migrated

9/11 repos in the named bucket have deno.json runtime config landed. The 2 holdouts (typed-wasm, ephapax) carry only vscode-extension package.json manifests, which fall under STEP 1's hypatia#391 path_allow_prefixes carve-out (vscode host-required class). No conversion needed.

Cross-cutting evidence

Estate-wide post-Layer-1 census:

• 177 total package.json post strict-excludes
• 508 total deno.json (Deno-native + shim co-existence pattern documented in STEP 2 MIGRATION.md)
• 74 directories have BOTH (Node/Bun interop shim alongside Deno-native config — legitimate per template Class B)
• 103 NPM_ONLY directories remain (the STEP 4/5/6 backlog)

Exemplar PR

panll#65 (merged 2026-05-30 18:55Z) is the canonical exemplar: package.json + package-lock.json deleted; deno.json adds nodeModulesDir: "auto" + npm:<pkg>@<semver> specifiers; CI workflows ported npm ci → deno install, npm run → deno task; full test suite identical to pre-migration.

Acceptance check

• ✅ One PR per repo across the bucket (mix of dedicated migration PRs + parallel deno.json adoptions during cross-cutting work)
• ✅ All merged with GREEN CI
• ✅ Hypatia scan estate-wide passes (STEP 1 carve-outs absorb the vscode-extension holdouts)
• ✅ Per-PR ownership gate verified

Closing as completed. STEPs 4-6 unblocked; STEP 7 closeout pending STEPs 4-6.

[hyperpolymath]

STEP 3 progress — 2026-05-30 (first batch of ≤2-manifest, root-level Class-A repos)

After STEP 1 (closed, hypatia rule + standards mirror) and STEP 2 (#315 ARMED, canonical template + 162/63 re-inventory), execution began on STEP 3 with the safest cohort: repos where deno.json already exists at root and package.json is either a redundant ReScript-toolchain stub or a script-delegation shim. All 5 below + 1 root-stub deletion shipped this session:

Repo	PR	Class	Status
odds-and-sods-package-manager	#42	A	auto-merge armed
avow-protocol	#18	A (root only — telegram-bot/ carve-out preserved)	auto-merge armed
coq-jr	#43	A (+ npx rescript → deno run npm:rescript fix per :npx_or_npm_run_in_ci)	auto-merge armed
blocky-writer	#12	A	auto-merge armed
rrecord-verity	#88	A (+ CI workflow swap deno/setup-deno)	MERGED
typed-wasm	#133	A (root stub only — tree-sitter-twasm/ carve-out)	auto-merge armed

Remaining STEP-3 backlog

49 repos total in the ≤2-manifest tier. After this batch, 43 remain. Triage shape from the next-cohort scan:

• ~12 repos are subdirectory manifests in carve-out classes (vscode-extension/, bindings/deno/, samples/, etc.) — should be skipped or path-prefix-allowlisted, not migrated. Will need a per-repo hypatia rule update PR rather than physical migration.
• ~15 repos are Class-B (npm-wrapper-via-Deno) for ReScript compiler — same shape as the canonical svalinn pattern (deno.json already has imports.rescript = "npm:rescript@...", tasks call rescript bare or via --node-modules-dir=auto). Each is a per-PR slice: convert task to deno run -A --node-modules-dir=auto npm:rescript + delete package.json stub. Exemplar: claude-integrations/gitlab-bridge.
• ~10 repos are heavier (frontend apps, polyglot tooling, MCP servers) — need real-per-repo judgment. Some are Class C (MCP host requires .js, npm-publish target).
• ~5 repos in developer-ecosystem/ are STEP 6's domain, not STEP 3.

Next sub-batch (Class B ReScript wrappers) is mechanically scriptable — recipe is in docs/migrations/npm-to-deno-template/MIGRATION.md (Class B section).

Closing this issue is gated on landing the remaining 43 repos' PRs. Keeping open as the STEP-3 tracker.

[hyperpolymath]

STEP 3 — second batch shipped (cumulative tally: 21 physical-migration PRs + 2 carve-out PRs)

Continuing from the 6-PR first batch comment above. The fan-out has now landed (auto-merge SQUASH armed on all) the next 11 PRs across two cohorts (Class-A redundant manifests + Class-B ReScript-toolchain wrappers), plus a 2-PR carve-out expansion that closes ~16 false-positive flags on root-level vscode-*, tree-sitter-*, /.lake/, /office-addin/, /bindings/{javascript,typescript}/ paths surfaced during triage.

Second Class-A/B physical-migration batch (11 new PRs)

Repo	PR	Class	Status
claude-integrations (gitlab-bridge/)	#50	B	armed
echidna (echidna-playground/)	#155	B	armed
rescript-ecosystem (bridge-web-rescript/)	#15	B	armed
zotero-tools (fogbinder/)	#27	B	armed
reposystem (recon-silly-ation/)	#88	B	armed
civic-connect (indieweb2-bastion/)	#51	B	armed
ambientops (broad-spectrum/)	#154	B	armed
no-nonsense-nntps (frontend/)	#25	B	armed
laniakea (client/)	#28	B	armed
flatracoon (netstack/interface/)	#21	B	armed
polyglot-i18n (src/)	#49	B	armed (also fixed bare-semver rescript imports → npm:rescript@^12.0.0)
gitbot-fleet (bots/sustainabot/bot-integration/)	#228	A	armed
social-media-polygraph (frontend/)	#35	B	armed
social-media-tools (polygraph/frontend/)	#38	B	armed
my-lang (frontier-practices/, vscode-extension carve-out preserved)	#82	B	armed

Carve-out expansion (Layer-1 gap-fix surfaced during STEP 3 triage)

Repo	PR	Purpose
hypatia	#405	Extends :nodejs_detected path_allow_prefixes with vscode-, tree-sitter-, /.lake/, /office-addin/, /bindings/{javascript,typescript}/
standards	#320	Mirrors hypatia#405 into .claude/CLAUDE.md §npm Exemptions table + rhodium-standard-repositories/.github/workflows/language-policy.yml grep -vE filters

These cover ~16 false-positive estate flags that otherwise would have been incorrectly migration-targeted: vscode-k9, vscode-a2ml, editors/vscode-007/, vscode-extension/ subdirs in empty-linter/error-lang/my-lang, tree-sitter-a2ml, tree-sitter-k9, editors/tree-sitter-ephapax, tree-sitter-affinescript/ subdirs in bofj-kitt/proof-burrower, proven/bindings/{javascript,typescript}/, absolute-zero/proofs/lean4/.lake/packages/, academic-workflow-suite/components/office-addin/.

Cumulative STEP 3 progress (since campaign open)

• 21 physical-migration PRs shipped/armed (1 MERGED, 20 auto-merge armed).
• 2 carve-out PRs shipped/armed (hypatia#405 + standards#320).
• STEP 1 [campaign #253] STEP 1 — LAYER-1 POLICY: hypatia path_allow_prefixes + standards mirror #261 CLOSED (substantively pre-shipped).
• STEP 2 [campaign #253] STEP 2 — OIKOS PATTERN AUDIT: canonical deno.json shape + re-inventory #262 closed by #315 (ARMED) — canonical template + 162/63 re-inventory.

Remaining STEP 3 backlog (~7-10 repos)

• Heavier Class B (full Vite/Vitest/Jest/Express stacks need real per-repo porting): excel-economic-numbers-tool, ubicity, anamnesis/visualization, git-scripts/ui, kaldor-iiot (backend+frontend), the-nash-equilibrium/mvt, patallm-gallery/claude-integrations/gitlab-bridge.
• Class B no-deno.json (need to author deno.json from template): safe-brute-force/rescript_modules, svalinn/ui.
• Deferred: panll (per memory — .claude/CLAUDE.md sweep restriction), misinformation-defence-platform (local git object corruption — needs fresh clone).
• Now-carve-out (no PR needed thanks to #405/docs(npm-exemptions): mirror hypatia#405 vscode-* + tree-sitter-* carve-outs (standards#253) #320): 007, absolute-zero, academic-workflow-suite, bofj-kitt, empty-linter, ephapax, error-lang, HOL, my-lang/vscode-extension, patallm-gallery/claude-integrations/firefox-lsp/vscode-extension, proof-burrower, proven, proven-servers, vscode-a2ml, vscode-k9.

Keeping this STEP-3 tracker open for the remaining ~7-10 heavier repos.

[hyperpolymath]

STEP 3 — final session tally 2026-05-30→31 (23 PRs LANDED)

All 22 STEP-3 physical-migration PRs + 1 Layer-1 carve-out expansion PR from this session have MERGED to main (mix of auto-merge SQUASH + admin-merge to clear queue). Local branches cleaned up; agent worktrees auto-pruned.

Landed (23 PRs across 22 repos)

Class A (pure-Deno port, just delete redundant manifest):
odds-and-sods-package-manager#42 · avow-protocol#18 · coq-jr#43 (+ npx rescript → deno run npm:rescript fix) · blocky-writer#12 · rrecord-verity#88 · typed-wasm#133 (root stub) · gitbot-fleet#228 (sustainabot/bot-integration/)

Class B (npm-wrapper-via-Deno, port rescript-toolchain wrappers):
claude-integrations#50 (gitlab-bridge/) · echidna#155 (echidna-playground/) · rescript-ecosystem#15 (bridge-web-rescript/) · zotero-tools#27 (fogbinder/) · reposystem#88 (recon-silly-ation/) · civic-connect#51 (indieweb2-bastion/) · ambientops#154 (broad-spectrum/) · no-nonsense-nntps#25 (frontend/) · laniakea#28 (client/) · flatracoon#21 (netstack/interface/) · polyglot-i18n#49 (src/ + legacy node.js.yml + test.yml deletions) · social-media-polygraph#35 (frontend/) · social-media-tools#38 (polygraph/frontend/) · my-lang#82 (frontier-practices/) · safe-brute-force#29 (rescript_modules/, fresh deno.json)

Layer-1 carve-out expansion (closes ~16 false-positive flags):
hypatia#405 — extends :nodejs_detected path_allow_prefixes with vscode-, tree-sitter-, /.lake/, /office-addin/, /bindings/{javascript,typescript}/. Mirrored by standards#320.

Source-fixed at CI for unblocking

polyglot-i18n#49 had 12 chronic CI failures from legacy upstream Node CI (node.js.yml Node 18-24 matrix + test.yml Node 10-20 × {ubuntu,windows,macos}). Both:

• ran npm ci + npm test against the root package.json (banned by cicd_rules/npx_or_npm_run_in_ci)
• had been red on main since the Deno migration began (last completed run = failure 2026-05-27)
• contradicted .claude/CLAUDE.md which explicitly bans Node + npm

Deleted at source in the same PR per the "resolve CICD at source before admin merge" directive.

Remaining STEP 3 backlog (~10 heavier repos)

Need real per-repo judgment:

• Full Vite/Vitest/Jest/Express stacks: anamnesis/visualization, git-scripts/ui, kaldor-iiot/backend/api (26 deps Express+jest), kaldor-iiot/frontend (26 deps Vite+eslint), patallm-gallery/claude-integrations/gitlab-bridge (18 deps tsx+vitest+tsc).
• Class C borderline: excel-economic-numbers-tool (Excel/LibreOffice add-in with @microsoft/office-js), ubicity (heavy node src/*.js invocations).
• Bun-specific: the-nash-equilibrium/mvt (bun test, bun src/server.js).
• Local-dep Class B: svalinn/ui (deps include cadre-router, rescript-tea — not npm-resolvable).
• Deferred:
  • panll (per memory: feedback_panll_claude_md_leave_alone.md — exclude from estate-wide sweeps).
  • misinformation-defence-platform/algorithm-shield (local git object corruption on the clone — needs fresh clone before any push).

Now-carve-out (no PR needed, covered by hypatia#405)

007, absolute-zero, academic-workflow-suite, bofj-kitt, empty-linter, ephapax, error-lang, HOL, my-lang/vscode-extension, patallm-gallery/claude-integrations/firefox-lsp/vscode-extension, proof-burrower, proven, proven-servers, vscode-a2ml, vscode-k9.

Standards-internal scaffolds (owner-managed, untouched)

standards/lol, standards/avow-protocol, standards/k9-svc/tools/vscode-extension (carve-out), standards/rhodium-standard-repositories/satellites/cccp/satellites/nextgen-languages/7-tentacles.

---

Keeping this issue OPEN as the STEP-3 longtail tracker. Acceptance will close when the remaining ~10 heavier repos land their per-repo migration PRs (each needs owner judgment on Class B vs C and whether the host contract requires staying on npm).