Phase ONS — Operator NEXT surfacing (portable + optional host niceties)

Status: Reviewed → pass (ONS-r2). ONS-a is spec-only and now frozen; no code lands in this phase. ONS-b (Auto) is cleared to build mechanically against this contract.

phase: ONS
outputs:
- id: ons-operator-next-surfacing
  path: docs/archive/phases/PHASE-ONS-OPERATOR-NEXT-SURFACING.md
  frozen: true
frozen_inputs:
- id: kh1-relay-standard
  path: docs/archive/phases/PHASE-KH1-HANDOVER-RELAY-STANDARD.md
- id: gs-paste-ready-regen
  path: docs/archive/phases/PHASE-GS-PASTE-READY-REGEN.md
- id: check-ok
  path: docs/archive/phases/PHASE-CHECK-OK.md
- id: k8-multi-lane
  path: docs/archive/phases/PHASE-K8-MULTI-LANE-DOCS-CONTRACT.md
- id: kit-spec-cli
  path: docs/OVERSEER-KIT-SPEC.md
- id: test-tiers
  path: policy/test-tiers.yaml
- id: model-labels
  path: policy/model-labels.yaml
- id: next-regen-extract
  path: tools/governance_hygiene/next_regen.py
- id: paste-block-extract
  path: tools/governance_hygiene/patch.py
- id: footprint
  path: cli/footprint.py
- id: cli-main
  path: cli/main.py
- id: handover-template
  path: templates/OVERSEER-HANDOVER.template.md
review_stamp:
  reviewed_at: '2026-08-12T17:43:55Z'
  verdict: pass
  reviewer_mode: agent
  reviewer_model: thinking-high
  reviewer_provider: local
  kit_version: 0.1.0
  artifact_digest: sha256:242e318feb880b5d632356e625f71a99f0a724dd5281afb80db2a0080fb463d8

Downstream edge: ONS-b treats this document as ground truth without re-deriving it (SPEC §6 mandatory reviewed freeze). It closes the operator-visible gap: disk handover/roadmap can be correct while an open IDE tab or an old chat paste still shows a stale wave. The kit cannot force any IDE to reload a tab. Master control is therefore a portable print contract (CLI + skill/rule + Copilot paste). Host niceties are optional and non-blocking.

Review record (§6.2): every freeze-review finding MUST cite file+line per SPEC §6; uncited findings are invalid and are discarded. Fixes during the loop are Tier 1 (feature branch); merge to main is Tier 3 and is never part of this loop.

Round	Reviewer	Verdict	Resolution
ONS-r1	Freeze-review loop (checklist + thinking, thinking-high)	findings	Semantic completeness: JSON lane null vs name; --print-next --lane compose; --print-next --all-lanes; heading_missing-before-extract; exact hook filenames; fifth skill grep (honesty). Fixed in-tree.
ONS-r2	Freeze-review loop (checklist + thinking, thinking-high)	pass	Checklist dry-run clean (0 findings). Semantic re-read: portable A1–A6 locked; niceties best-effort and non-DONE-gate; GS-PASTE regen not reopened; ok next ≠ ok workspace check-next; exit 37; no tab-reload claim; no per-branch handover names. Stamp written by ok review --freeze.

---

§ONS.0 — Simple summary

Agents already update the living handover and roadmap on disk. Operators often keep an old editor tab or an old chat paste open, so they see yesterday’s next step even when the files on disk are current. This phase freezes a portable way to print the current paste-ready prompt from disk (ok next), and a rule that the agent’s last message after updating those files must include that same printed fence. Optional editor extras may make a tab more likely to refresh. They do not guarantee the open tab is right, and they are not a gate for DONE.

Technical summary: add a read-only CLI ok next (synonym: ok governance-sync --print-next) that extracts the KH1 paste-ready fence from the config-driven handover path and prints it under a fixed heading. Fail closed if the file or fence is missing/malformed. Vendor a skill + alwaysApply rule to .cursor/ and .claude/ via existing ok sync footprint globs. Ship Copilot/other-assistant instruction text (docs/PRINT-NEXT.md + AGENTS/install notes). Optional Cursor stop hook template is opt-in, fail-open, and must not replace the portable contract. Reuse extract_paste_fence_body (tools/governance_hygiene/next_regen.py). Do not regenerate NEXT (GS-PASTE remains the sole regen surface). Do not add per-branch handover filenames.

---

§ONS.1 — Verified problem (do not redesign)

Fact	Evidence
One living handover filename per repo, config-driven	.overseer/config.yaml docs.handover; cli/docs_paths.py living_doc_abs / join_docs_rel
Closeout already requires roadmap + handover on the feature branch	SD-17; cursor/rules/governance-sync.mdc; KH1
Paste fence already has an extractor	extract_paste_fence_body in tools/governance_hygiene/next_regen.py (regex _PASTE_FENCE_RE); extract_paste_ready_block in tools/governance_hygiene/patch.py
KH1 requires ### Paste-ready prompt + Model: in the fence	docs/archive/phases/PHASE-KH1-HANDOVER-RELAY-STANDARD.md H7 / H8
GS-PASTE regenerates those anchors via ok governance-sync only	docs/archive/phases/PHASE-GS-PASTE-READY-REGEN.md §GSP.3
GS-PASTE forbade a regen subcommand	same §GSP.1 / §GSP.3 “Any other command MUST NOT regenerate NEXT / paste”
Skills vendor to Cursor and Claude Code; Copilot uses CLI + paste doc	cli/footprint.py dual dest; docs/archive/phases/PHASE-CHECK-OK.md; docs/CHECK-OK.md
Automations/hooks are templates, not auto-enabled	cursor/README.md; GFG session-end Automation is Tier 2 confirm-once
ok workspace check-next already exists (constellation relays, exit 35)	SPEC §5; cli/main.py workspace subparser
Kit cannot force IDE tab reload	Operator problem statement; no existing CLI/API in this repo reloads editor buffers

---

§ONS.2 — Scope

In scope (ONS-a freezes; ONS-b implements):

1. Portable contract A1–A6 (§ONS.3–§ONS.8).
2. Optional host niceties B (§ONS.9) — documented as best-effort; must not replace A.
3. Seven-tier matrix (§ONS.12) and Definition of Done (§ONS.13).
4. Additive SPEC §5 row, COMMANDS membership, kit AGENTS / install-doc text.

Out of scope (explicit non-goals):

Non-goal	Why rejected
Claiming the kit can force IDE tab reload	False assurance; CLI/Muse/shell rewrites and dirty buffers can still stale a tab
Per-branch handover file names	Violates A1; one config-driven living path
Requiring merge to main to “refresh” NEXT	NEXT lives on the feature branch (SD-17); merge is Tier 3
Making tab sync a gate for DONE	Niceties are non-blocking; BV + tests + docs remain the DONE gates
Regenerating NEXT / paste via ok next	GS-PASTE sole regen surface; this command is read-only extract
Redesigning product waves / KH1 shape / GS-PASTE selector	Operator: do not redesign
Aliasing ok next to ok workspace check-next	Different command; constellation vs paste-print
ONS-b Auto implementation in the Thinking phase	SD-3 split
Live consumer ok sync / re-init	Operator-gated
Native Copilot Agent Skills	GitHub does not support SKILL.md; CLI + paste is the Copilot contract (Check OK precedent)
Auto-enabling Cursor hooks/Automations on ok sync	Tier 2 confirm-once; same as GFG

GS-PASTE supersession (narrow): docs/archive/phases/PHASE-GS-PASTE-READY-REGEN.md §GSP.1 rejected a “Separate CLI / slash command for NEXT regen.” This phase does not reopen regen. It adds a read-only print command. §GSP.3 “Any other command MUST NOT regenerate NEXT / paste” remains in force for ok next and for ok governance-sync --print-next.

---

§ONS.3 — A1. One living handover path (invariant)

1. The living handover path is only the config-driven destination: join_docs_rel(config.repo.root_relative_docs, handover_name) where handover_name is config.docs.handover or, when docs.lanes is set, the selected lane’s handover (resolve_lane_docs, §ONS.5.8).
2. ONS-b MUST NOT introduce per-branch handover filenames, per-wave copies, or OVERSEER-HANDOVER-<branch>.md patterns.
3. K13 {REPO_SLUG}-OVERSEER-HANDOVER.md board identity (§MR.6.5) is unchanged — that is a repo-slug prefix when workspace: is configured, not a branch suffix.

---

§ONS.4 — A2. Closeout still updates both docs (existing SD-17)

ONS does not change when or where closeout writes. Agents still update {{docs.roadmap_path}} and {{docs.handover_path}} together on the feature branch. ok next runs after those bytes are on disk. A session that updates one living doc without the other remains incomplete under existing SD-17.

---

§ONS.5 — A3. CLI: ok next (canonical) + --print-next (synonym)

§ONS.5.1 — Surfaces

Surface	Behavior
ok next	Canonical. Read-only. Print the current paste-ready fence from disk.
ok governance-sync --print-next	Synonym. Short-circuits to the same extractor. Does not run R1–R5, D1–D4, patches, commits, or marker stamps.
ok governance-sync --print-next --write	Fail closed — mutually exclusive. Exit 2. Message must include print-next mutually exclusive with --write.
ok governance-sync --print-next --all-lanes	Fail closed — mutually exclusive. Exit 2. Message must include print-next mutually exclusive with --all-lanes. One fence per invocation.
ok governance-sync --print-next --lane NAME	Allowed. Same resolve_lane_docs as ok next --lane.
ok governance-sync --print-next --dry-run	Allowed; --print-next still short-circuits (print-only). --dry-run does not add a plan.
Any other command	Must not print this heading as a substitute for ok next (skills may instruct agents to run ok next).

ok next MUST NOT call muse or git. Config load + filesystem read only.

§ONS.5.2 — Module + wiring (exact paths)

Piece	Path
Extract + format	tools/print_next/ (extract_current_next, format_current_next, CurrentNextResult)
Reuse	extract_paste_fence_body from tools/governance_hygiene/next_regen.py — do not fork the regex
CLI command	cli/commands/next.py (run_next_command)
Parser	cli/main.py: add "next" to COMMANDS; subparser next; governance-sync flag --print-next
Dispatch	args.command == "next" → run_next_command; --print-next on governance-sync → same function, skip run_governance_sync

§ONS.5.3 — Frozen heading

Exact heading string (one space before the Unicode em dash —, matching the operator prompt byte-for-byte):

## CURRENT NEXT — paste this

Constant name in code: CURRENT_NEXT_HEADING. Tests assert this exact string.

§ONS.5.4 — Human stdout (exit 0)

Print exactly these bytes (UTF-8), with a trailing newline on the last line:

1. CURRENT_NEXT_HEADING
2. one newline
3. one newline (blank line)
4. opening fence line ```text
5. newline
6. fence body as returned by extract_paste_fence_body (no extra strip beyond what that function already captures)
7. if the captured body does not already end with a newline, append one newline
8. closing fence line ```
9. newline

Do not prepend ### Paste-ready prompt. That heading stays in the handover file. The chat/CLI surface uses CURRENT_NEXT_HEADING plus the fence body.

--quiet still prints this block (the block is the product). Diagnostics stay on stderr.

§ONS.5.5 — --json stdout (exit 0)

Single JSON object, no human wrap:

{
  "ok": true,
  "path": "<repo-relative POSIX handover path>",
  "lane": "<lane id or null>",
  "heading": "## CURRENT NEXT — paste this",
  "fence": "<body>",
  "error": null
}

On success, lane is JSON null when docs.lanes is unset; otherwise the resolved lane name string (--lane or docs.default_lane).

On failure with --json: ok: false, fence: null, error = closed reason token (§ONS.5.7), message = human string. Exit code still §ONS.5.6 (not 0).

§ONS.5.6 — Exit codes

Code	When
0	Fence extracted and printed
2	Config error; --print-next combined with --write or --all-lanes; --lane invalid / not configured (same class as other CLI config refusals)
4	Config path outside repo root (existing is_within_repo refusal)
37	EXIT_NEXT_MALFORMED — handover missing/unreadable or fence missing/malformed (§ONS.5.7)

Do not reuse 35 (ok workspace check-next) or 36 (EXIT_POST_LAND_SYNC). Unknown command remains 1.

Constant: EXIT_NEXT_MALFORMED = 37 in cli/commands/next.py (and imported by tests).

§ONS.5.7 — Fail-closed reasons (closed vocabulary)

stderr (non-JSON) prefix: next: <reason> — <detail>

reason	Trigger
handover_missing	Resolved handover path does not exist on disk
handover_unreadable	Exists but cannot be read as UTF-8 text
heading_missing	Handover text has no ### Paste-ready prompt heading (KH1 H7). Checked before fence extract so a stray triple-backtick elsewhere cannot become the paste fence.
fence_missing	Heading present but extract_paste_fence_body returns None
fence_empty	Body after str.strip() is empty
model_missing	Body does not contain the substring Model: (KH1 H8)

First matching reason in the order above wins. Do not invent a fence from roadmap glance, change-log, chat memory, or next_regen planned bytes.

§ONS.5.8 — Flags

ok next:

Flag	Meaning
--lane NAME	Use that docs.lanes handover. Absent docs.lanes or unknown name → exit 2.
(globals)	-C/--repo, --config, --json, -q/--quiet, -v/--verbose, --no-color as existing CLI

No --write. No --apply. Resolve the handover file with existing adapters.config.resolve_lane_docs(config, lane) (adapters/config.py): --lane omitted uses docs.default_lane when docs.lanes is set; --lane when docs.lanes is absent raises ConfigError (CLI exit 2); unknown lane name raises ConfigError (exit 2). Do not fork a second default-lane rule.

---

§ONS.6 — A4. Skill + alwaysApply rule (Cursor and Claude Code)

§ONS.6.1 — Deliverable paths

Item	Kit source	Vendored dest (existing footprint globs)
Skill	cursor/skills/print-next/SKILL.md	.cursor/skills/print-next/SKILL.md and .claude/skills/print-next/SKILL.md
Rule	cursor/rules/print-next-closeout.mdc	.cursor/rules/print-next-closeout.mdc (alwaysApply: true)

cli/footprint.py already copies every cursor/rules/* file and every cursor/skills/** file to both skill trees. ONS-b adds the files; it does not change glob logic unless a test proves a new file is skipped (then fix the glob — do not special-case).

§ONS.6.2 — Frozen skill/rule obligations (normative text)

Both files MUST contain all of the following substrings (tests grep them):

1. ## CURRENT NEXT — paste this
2. ok next
3. read from disk after write
4. Session incomplete without it
5. do not guarantee an accurate open tab

Normative behavior (write this into the skill and the alwaysApply rule):

After any update to the living handover and/or roadmap, the agent’s FINAL reply MUST include the full paste-ready fence under heading ## CURRENT NEXT — paste this. Bytes MUST come from disk after the write (run ok next, or Read the handover file and extract the ### Paste-ready prompt fence). Do not paste from memory, from an earlier chat, or from an unsaved editor buffer. A session that updated those docs and omits this block is incomplete.

Prefer native editor tools (Cursor StrReplace/Write; Claude Code Edit) for handover/roadmap so the host is more likely to refresh an open tab. That preference is best-effort and does not replace ok next.

§ONS.6.3 — Also touch (additive one-liners, not a second protocol)

File	Additive requirement
cursor/rules/governance-sync.mdc	Closeout list item: after writing both docs, print ok next / CURRENT NEXT block
cursor/rules/orchestrator.mdc	Starting-a-phase / close step: final reply includes CURRENT NEXT from disk
cursor/skills/governance-sync/SKILL.md	After apply/write of living docs, run ok next (or equivalent Read) and include the heading+fence in the user-visible close
cursor/README.md	Table rows for the new skill + rule; Copilot fallback ok next + docs/PRINT-NEXT.md
templates/OVERSEER-HANDOVER.template.md	Regeneration-rules list item: print CURRENT NEXT from disk after closeout

---

§ONS.7 — A5. Copilot / other assistants

Item	Path	Role
Paste prompt	docs/PRINT-NEXT.md	Same job as docs/CHECK-OK.md: CLI + paste for tools without SKILL.md
Kit agent doc	AGENTS.md	Short “Print NEXT on closeout” section: prefer ok next; never trust the open tab
Consumer stubs	docs/consumers/scooling/OVERSEER-SETUP.md, docs/consumers/knowtation/OVERSEER-SETUP.md, docs/consumers/videofactory/OVERSEER-SETUP.md	One paragraph each: after handover/roadmap update, print fence via ok next
Cursor README	cursor/README.md	Already listed in §ONS.6.3

docs/PRINT-NEXT.md MUST include:

• Fast path: ok next (and ok governance-sync --print-next)
• The heading ## CURRENT NEXT — paste this
• Source-of-truth order (§ONS.8)
• Runtime map table (Cursor / Claude Code / Copilot) matching Check OK’s shape
• Explicit: niceties do not guarantee tab accuracy

ok sync already ships the alwaysApply rule to Cursor. Copilot does not load .mdc. Install docs + docs/PRINT-NEXT.md + AGENTS.md are the Copilot contract. Do not vendor a Copilot-only hook.

---

§ONS.8 — A6. Operator source of truth order (frozen)

When the printed fence, the disk file, and an open tab disagree:

1. ok next / the printed CURRENT NEXT fence (this session, after the write)
2. The handover file on disk at the config-driven path
3. Never “whatever the open tab happens to show”

An old chat paste is not a source of truth. Merge to main is not required to refresh NEXT.

---

§ONS.9 — B. Optional host niceties (non-blocking)

Niceties improve odds of tab refresh. They do not guarantee an accurate open tab. CLI, Muse, shell redirects, and dirty buffers can still leave a stale tab. Niceties MUST NOT replace §ONS.5–§ONS.8. Tab sync MUST NOT be a DONE gate.

§ONS.9.1 — Cursor

Nicety	Frozen HOW
Prefer native edit tools	Skill/rule text only (§ONS.6.2)
Optional stop hook	Kit template under cursor/hooks/ — not auto-enabled; not added to resolve_footprint in this phase (same posture as cursor/automations/)
Tab sync	Best-effort only; no API claim

Stop-hook template files (exact; ONS-b must not invent extra names):

Path	Contents
cursor/hooks/README.md	Enable instructions; honesty sentence; Tier 2 confirm-once; not in footprint
cursor/hooks/print-next-stop.json	A snippet to merge into a project .cursor/hooks.json — not a full overwrite of consumer hooks

Stop-hook snippet requirements:

• Event: stop
• failClosed: false (fail open)
• loop_limit: 1 (at most one follow-up)
• If the agent’s last reply lacks ## CURRENT NEXT — paste this and this session wrote the living handover or roadmap, follow up once: run or request ok next and include the heading+fence
• MUST NOT block DONE, MUST NOT merge, MUST NOT write docs, MUST NOT claim tab reload
• Operator enable = Tier 2 confirm-once (document in cursor/README.md)

If the host stop payload cannot see “wrote handover,” the template follow-up may fire whenever the heading is absent. That is acceptable because loop_limit is 1 and fail-open. Do not invent a file-watcher.

§ONS.9.2 — Claude Code

Same skill bytes as Cursor (already dual-vendored). Prefer native Edit for living docs. No Cursor hooks.json in Claude Code. Best-effort tab sync only. No additional Claude-only hook required in ONS-b.

§ONS.9.3 — Copilot

Instruction text only (docs/PRINT-NEXT.md, AGENTS.md, consumer stubs). No Cursor hooks. Best-effort when Copilot’s editor applies edits.

§ONS.9.4 — Honesty sentence (must appear in freeze, skill, PRINT-NEXT, cursor README)

Host niceties improve odds of tab refresh; they do not guarantee an accurate open tab. CLI/Muse/shell rewrites and dirty buffers can still stale a tab.

---

§ONS.10 — Name collision table

Command	Meaning	Exit on fail
ok next	Print paste-ready fence from disk (this phase)	37 malformed; 2 config
ok governance-sync --print-next	Synonym for ok next	same
ok workspace check-next	Constellation relay freshness (K13)	35
ok governance-sync (no --print-next)	Hygiene agent; may regen NEXT (GS-PASTE)	existing

Help text for ok next MUST include: Print the paste-ready NEXT fence from disk (read-only). Not workspace check-next.

---

§ONS.11 — Security / privacy

• Handover text is data — never interpolate into a shell command line.
• ok next is read-only; no commits; no network; no Muse; no git.
• Do not print secrets if a handover illegally contains them (no extra redaction engine in this phase; existing “no secrets in governance docs” policy stands). Tests include a fixture fence with a fake sk- token and assert the CLI still prints it as data (it must not execute it) and does not add it to argv.
• Path confinement: handover path must resolve under repo_root (is_within_repo).
• --lane values are identifiers, not path fragments; reject .. and slashes in lane names with exit 2.
• git-only fixtures: zero muse argv (vacuous: command never calls VCS).

---

§ONS.12 — Seven-tier test matrix (ONS-b)

Tier	Frozen case
unit	(1) Valid KH1 handover → heading + body; (2) missing file → handover_missing / 37; (3) no ### Paste-ready prompt → heading_missing / 37 even if another ``` fence exists; (4) heading present but no fence → fence_missing / 37; (5) empty body → fence_empty / 37; (6) no Model: → model_missing / 37; (7) CURRENT_NEXT_HEADING exact; (8) extract_paste_fence_body reused (body matches helper on same text)
integration	ok next on config-git-only fixture: exit 0; stdout contains heading + Model:; runner log contains no muse and no git; ok governance-sync --print-next same stdout/exit; --print-next --write exit 2; --print-next --all-lanes exit 2
e2e	Write a new fence to the fixture handover, then ok next prints the new body (not the previous one). main untouched. Second ok next identical (idempotent read).
stress	1.5 MiB handover with the paste fence at the end extracts under the performance bound; 200 DONE rows elsewhere do not change the extracted body
data-integrity	ok next twice → identical stdout; command does not modify handover/roadmap/lock bytes (hash before/after)
performance	Completes under 2.0s on kit-sized handover in the existing test harness (no unbounded docs/ walk)
security	Path escape (--config outside repo) → 4; lane name ../x → 2; git-only fixture zero Muse argv; skill+rule files contain the five required substrings (§ONS.6.2); docs/PRINT-NEXT.md contains the honesty sentence (§ONS.9.4); cursor/hooks/print-next-stop.json has failClosed false

---

§ONS.13 — Definition of Done (ONS-b)

• [ ] ok next and ok governance-sync --print-next implement §ONS.5
• [ ] Fail-closed reasons + exits 0/2/4/37 as frozen
• [ ] Skill + alwaysApply rule vendored to .cursor/ and .claude/ via ok sync globs
• [ ] Copilot path: docs/PRINT-NEXT.md + AGENTS + consumer-stub paragraph
• [ ] Host niceties documented as best-effort; optional hook not auto-enabled; honesty sentence present
• [ ] SPEC §5 additive ok next row; COMMANDS includes next
• [ ] Seven-tier §ONS.12 green locally; git-only fixture proves no Muse
• [ ] /build-verification-review → pass before ROADMAP ONS-b DONE
• [ ] docs/ROADMAP.md + docs/OVERSEER-HANDOVER.md updated together (SD-17)
• [ ] No consumer re-init; no feature→GitHub-main; no secrets
• [ ] ONS-b Auto does not mark tab sync as a DONE gate

---

§ONS.14 — Hard stops

• No ONS-b Auto implementation during ONS-a
• No merge to main / staging push / live posture flips without Tier 3
• No live consumer re-init
• No inventing a paste fence when extraction fails
• No claiming IDE tab reload
• No per-branch handover filenames
• No Muse required for baseline green

---

§ONS.15 — Auto paste (ONS-b) — frozen for handover after ONS-a pass

You are Auto on overseer-kit — ONS-b Operator NEXT surfacing.

Model: Auto
Repo: overseer-kit
Branch: feat/ons-operator-next-surfacing
Step: ONS-b
Authority: authoritative

Read first (frozen; do not redesign):
- docs/archive/phases/PHASE-ONS-OPERATOR-NEXT-SURFACING.md
- docs/ROADMAP.md row ONS-b
- docs/OVERSEER-HANDOVER.md

Build ONS-b and ONLY ONS-b:
- tools/print_next/ + cli/commands/next.py + ok next + governance-sync --print-next
- cursor/skills/print-next/SKILL.md + cursor/rules/print-next-closeout.mdc
- docs/PRINT-NEXT.md + AGENTS.md + consumer stub paragraphs + SPEC §5 row
- optional cursor/hooks/ template (not footprint, failClosed false)
- seven-tier §ONS.12; /build-verification-review → pass
- update ROADMAP ONS-b + handover; feature-branch commit

Do NOT: regenerate NEXT (GS-PASTE); alias workspace check-next; auto-enable hooks;
claim tab reload; merge to main; live consumer sync; start other phases.

Model: Auto

---

§ONS.16 — Cross-references

• docs/archive/phases/PHASE-KH1-HANDOVER-RELAY-STANDARD.md — H7/H8 paste shape
• docs/archive/phases/PHASE-GS-PASTE-READY-REGEN.md — sole regen surface
• docs/archive/phases/PHASE-CHECK-OK.md — dual skill vendor + Copilot paste doc
• docs/archive/phases/PHASE-K8-MULTI-LANE-DOCS-CONTRACT.md — --lane
• docs/OVERSEER-KIT-SPEC.md §5 — CLI table (ONS-b additive row)
• policy/test-tiers.yaml — seven tiers
• cli/footprint.py — rules + dual skills glob
• tools/governance_hygiene/next_regen.py — extract_paste_fence_body