@event4u/agent-config 10.4.0 → 11.0.0

package/.claude-plugin/marketplace.json

@@ -6,7 +6,7 @@
   },
   "metadata": {
     "description": "Bootstrap shim for the event4u agent-config suite. This plugin ships ONLY the dispatcher hooks plus one install-pointer skill — ALL content (skills, rules, commands) is installed via `npx -y @event4u/agent-config init`, which writes the file projection AND registers the same hooks in ~/.claude/settings.json. DEPRECATED as a content channel; kept so the marketplace listing stays discoverable and existing installs keep working hooks.",
-    "version": "10.4.0",
+    "version": "11.0.0",
     "keywords": [
       "agent-config",
       "skills",

package/CHANGELOG.md

@@ -348,6 +348,75 @@ Entry-shape contract: [`docs/contracts/CHANGELOG-conventions.md`](docs/contracts
 > that forces a new era split (`# Era: 10.4.x`, etc.) — see
 > [`docs/contracts/CHANGELOG-conventions.md § Era splits`](docs/contracts/CHANGELOG-conventions.md).
 
+## [11.0.0](https://github.com/event4u-app/agent-config/compare/10.4.0...11.0.0) (2026-08-13)
+
+### Release highlights
+
+<!-- Curated head: fill before merge, keep it under 10 lines, and leave `_none_` where it is genuinely the answer. The generated log below is unchanged. -->
+- **Behaviour changes:** _auto-derived, rewrite before merge:_ rule/schema diffs, breaking commits or removed public surface in 7d4af2c, 9e4cc35.
+- **Default changes + migration:** _none_
+- **Security and correctness:** _none_
+- **Honest nulls:** _none_
+- **Known limitations:** _none_
+
+### BREAKING CHANGES
+
+* **settings:** delete worktrees.mode — worktree creation is instruction-only ([9e4cc35](https://github.com/event4u-app/agent-config/commit/9e4cc35513a69b9e107b639ff9dc78622d42b09e))
+
+### Features
+
+* **council:** replace the per-run spend confirmation with a ceiling ([945d6c0](https://github.com/event4u-app/agent-config/commit/945d6c08b6e9a5519858fa6f4bb6ddece461621b))
+* **gates:** resolve the PR base before judging branch freshness ([f77da9a](https://github.com/event4u-app/agent-config/commit/f77da9a6618d0ef982cd4b9bb83201179e26e818))
+* **grounding:** the three 1-10 design dials, re-derived not ported ([1516fa8](https://github.com/event4u-app/agent-config/commit/1516fa8adda1c772a71062fb1950598f336b1c5e))
+* **corpus:** vendor motion.csv and a gsap search domain; decline the rest ([0a82fa2](https://github.com/event4u-app/agent-config/commit/0a82fa2128ffdbc3273b85efd83152ead8e1da28))
+
+### Bug Fixes
+
+* **settings:** correct the class-count summaries after the deletion ([b1cf141](https://github.com/event4u-app/agent-config/commit/b1cf1419366bd1a8bc338ce4a11e310724143627))
+* **session-register:** gate the context block on a real collision ([c1dcf0e](https://github.com/event4u-app/agent-config/commit/c1dcf0e599e709f465903805a3f8147cc079fecd))
+* **council:** align the command surfaces with the ceiling bound ([2e3ad3f](https://github.com/event4u-app/agent-config/commit/2e3ad3f74947fba2ff73c2017515e5954b19cefb))
+* **session-register:** anchor the record on the session checkout, not the chdir target ([7b64d4e](https://github.com/event4u-app/agent-config/commit/7b64d4e3d0db42e6f177692cdf2c47eba3634d72))
+* **review:** put the skip declaration on one line ([d6d3736](https://github.com/event4u-app/agent-config/commit/d6d3736fef1509eda589dbb8f36a3b806f30bcf3))
+* **docs:** drop maintainer-only invocations from the shipped surfaces ([4e89881](https://github.com/event4u-app/agent-config/commit/4e8988172a773be43b8a9b3d93d3676f76fc5adf))
+* **roadmap:** relocate distillation's two maintainer-gated items and archive it ([c19fcf6](https://github.com/event4u-app/agent-config/commit/c19fcf6072f499a5e0b5149985e238ee8eb5511e))
+* **corpus:** advance last_checked in the manifest too, not only in ATTRIBUTION ([ca40b92](https://github.com/event4u-app/agent-config/commit/ca40b92b83db4a1951b4dc1328b21ecfdf7919f4))
+* **provenance:** correct a falsified compliance record, and shrink two diffs ([88e095d](https://github.com/event4u-app/agent-config/commit/88e095d8342106111bd36d69f410f39646deb558))
+* **grounding:** make the dials reach a reader, and make the variance claim true ([af919d3](https://github.com/event4u-app/agent-config/commit/af919d30d91a2c379819135fa7e34446a7da33a3))
+* **gates:** carve out the two license-attribution surfaces for code borrows ([2a969a0](https://github.com/event4u-app/agent-config/commit/2a969a0c4835e9a4db031d01e70ebcb5e0bb051d))
+* **records:** close all seven R2 findings, including a false claim of mine ([5ca4290](https://github.com/event4u-app/agent-config/commit/5ca4290795560d4fca8a054db6439e39e5bb6b42))
+
+### Documentation
+
+* **skills:** hardcode the instruction-only worktree rule across every surface ([7d4af2c](https://github.com/event4u-app/agent-config/commit/7d4af2c9b51714fe3a6ae636b379f835f7dc0cbc))
+* **adr:** record ADR-229 — worktree creation is instruction-only ([b20e3a5](https://github.com/event4u-app/agent-config/commit/b20e3a56a05402d95f88a66ade4075b3d211e53d))
+* **adr:** record ADR-230 — the council spend bound is a ceiling ([1f35330](https://github.com/event4u-app/agent-config/commit/1f35330a6203c336a9ddced6593cf7f8e09fdbdc))
+* **review:** declare the completion-review skip for a docs-only branch ([3ad11bf](https://github.com/event4u-app/agent-config/commit/3ad11bf5109088bca10a80954dffbd280e2586c1))
+* **roadmap:** close Phase 0 Step 3 and re-scope Phases 1 and 2 on the measurement ([d9a888e](https://github.com/event4u-app/agent-config/commit/d9a888e793977be05af7c2c2b345655956e3eb8e))
+* **evidence:** reproduce #58109 and record what the ledger cannot see ([21b8d2d](https://github.com/event4u-app/agent-config/commit/21b8d2dfbcd0dff415bfda7306788a86f3b3809b))
+* **review:** re-bind the skip declaration after merging main ([3e03c75](https://github.com/event4u-app/agent-config/commit/3e03c75c4465ceac9d02bdfb76e8ad0091387965))
+* **git:** route the pre-PR freshness check through the gate ([c0edb1f](https://github.com/event4u-app/agent-config/commit/c0edb1f36c001087d1118ef1fac3e7cb5e10f333))
+* **review:** re-scope the skip declaration after the main merge ([7184ae8](https://github.com/event4u-app/agent-config/commit/7184ae89ec9e9385aca59ecba9bdd2eacc336c08))
+* **review:** re-bind the findings artefact after the corpus-staleness fix ([47131d4](https://github.com/event4u-app/agent-config/commit/47131d403539cdaebf67e7b514267562835ae5c1))
+* **review:** re-bind the findings artefact to the post-fix scope ([6a48f11](https://github.com/event4u-app/agent-config/commit/6a48f11ba78b1086209dfa06928bfe1f7addc743))
+* **roadmap:** correct three overclaims, re-depth the archive links, stale watchlist ([c4e95d3](https://github.com/event4u-app/agent-config/commit/c4e95d36a6efcaf44f1aa693816806111e823165))
+* **review:** bind the R2 findings for feat-design-system-onramp-blockers ([2092734](https://github.com/event4u-app/agent-config/commit/2092734d3438ac2647bdcdde5faf05be49b7b0dc))
+* **review:** re-bind the artefact after the third main merge ([3b4bdb6](https://github.com/event4u-app/agent-config/commit/3b4bdb6a08f46181d869a2d4bd4dfb16af3e4349))
+* **review:** re-bind the R2 artefact after the fix pass ([5bfc84c](https://github.com/event4u-app/agent-config/commit/5bfc84c8bfea61a9522f21f8cbb7f72acdcc7450))
+* **council:** record the two unblocking decisions, and their honest standing ([8cd7182](https://github.com/event4u-app/agent-config/commit/8cd71823c1b1ce3c6002d16cb444d1b96b70f3fa))
+* **review:** the real R2 review, replacing the skip — 7 findings ([834e7f1](https://github.com/event4u-app/agent-config/commit/834e7f10f26e0d93911f0c92d0720f72c8ee0fc2))
+
+### Tests
+
+* **settings:** re-point two vacuous assertions and assert the deletion ([933f320](https://github.com/event4u-app/agent-config/commit/933f3208efd38a22854e147e7d3c7ddee2bacde1))
+
+### Chores
+
+* **roadmap:** regenerate the dashboard after the main merge ([2653222](https://github.com/event4u-app/agent-config/commit/26532222c07d0784f1f608a0fc4185ed9b02274b))
+* **roadmap:** regenerate the dashboard after the third main merge ([b271d07](https://github.com/event4u-app/agent-config/commit/b271d078b70d3c177571fc152cc00315b5780aba))
+* **roadmap:** close design-system-onramp at 12/12 and archive it ([14bce9c](https://github.com/event4u-app/agent-config/commit/14bce9c59b193631737f7b2a19157fd53d4e6de4))
+
+Tests: 13818 (+50 since 10.4.0)
+
 ## [10.4.0](https://github.com/event4u-app/agent-config/compare/10.3.0...10.4.0) (2026-08-13)
 
 ### Release highlights

package/dist/agent-src/commands/council/debate.md

@@ -68,11 +68,20 @@ this automatically when invoked via `debate`). The estimate covers
 projection. Surface the projection to the user with a clear note that
 progressive disclosure may stop the debate early.
 
-### 4. Cost confirmation (ALWAYS ASK)
+### 4. Spend bound (ask when a billable member has no ceiling)
 
-Per [`ai-council` skill § cost gate](../../skills/ai-council/SKILL.md),
-always surface the projected total spend and ask for explicit consent
-before invoking. Numbered options:
+Per [`ai-council` skill § Procedure 3](../../skills/ai-council/SKILL.md)
+and ADR-230, surface the projected total spend in every case, and ask
+only when a billable member is in the fan-out AND both
+`cost_budget.max_total_usd` and `daily_limit_usd` are `0`.
+
+A debate multiplies the per-round cost by `N`, so quote the worst case,
+not one round — that is what a ceiling has to cover. The between-rounds
+confirmation gate is unaffected: it is a separate, per-round control and
+it still fires, which is where a debate that turns out expensive gets
+stopped.
+
+Numbered options, when the ask applies:
 
 > The debate is projected to cost **$X.XXXX** ($Y.YYYY × N rounds, worst case).
 >

package/dist/agent-src/commands/council/default.md

@@ -152,11 +152,22 @@ Construct each member from the resolved mode:
 - `manual` → `ManualClient` from `scripts.ai_council.clients`
   (`billable=False`, no API key, no SDK call).
 
-### 3. Cost confirmation — ALWAYS ASK for billable members
+### 3. Spend bound — ASK when a billable member has no ceiling
 
-Council calls to billable members spend money. Even under
-`personal.autonomy: on`, the agent **must** ask before invoking any
-billable member.
+Council calls to billable members spend money, so a bound is mandatory.
+Per ADR-230 the bound is a **ceiling**, not a per-invocation approval —
+the same three cases the `ai-council` skill resolves in its Procedure § 3:
+
+| Situation | Behaviour |
+|---|---|
+| No billable member (all `mode: cli`, subscription auth) | No gate. Estimate is information; go to Step 4. Spend is $0. |
+| Billable member **with** `cost_budget.max_total_usd` or `daily_limit_usd` non-zero | No per-run ask. The ceiling is the authorization the user already gave; `on_overrun` still pauses per member on breach. |
+| Billable member with **both** caps `0` | **Ask.** Nothing bounds the spend, so the user must. |
+
+`personal.autonomy` neither creates nor lifts that bound — autonomy is
+not a ceiling. Consumers who want a per-invocation gate set a small
+`cost_budget.max_total_usd`; every call then breaches and `on_overrun`
+asks per member.
 
 Run the CLI in **estimate** mode first — it bundles the artefact, runs
 redaction, and prints the per-member preview without spending:
@@ -182,17 +193,21 @@ roadmap file with `--input-mode roadmap`. `diff` and `files` modes
 remain Phase 4 — for now ask the user to convert into a `prompt`.
 
 The CLI prints a `council:estimate · members=N (billable=M)` line
-followed by per-member projected USD and a TOTAL. Render that to the
-user inside the cost-confirmation numbered-options block per the
-`ai-council` skill (§ Pre-call estimate format) — then `1. Run /
-2. Cancel`. If the billable count is `0`, skip the gate entirely
-(spend = $0) and proceed directly to Step 4.
+followed by per-member projected USD and a TOTAL. Render that table per
+the `ai-council` skill (§ Pre-call estimate format) in every case — it is
+what makes the spend legible whether or not it is being asked about.
+
+Then read `billable=M` together with the two caps:
 
-Wait for the user's pick. `1` proceeds; anything else aborts.
+- `M == 0` → no gate. Proceed to Step 4.
+- `M > 0` and a cap is set → no gate. Proceed to Step 4; `on_overrun`
+  handles a breach per member.
+- `M > 0` and both caps are `0` → render the table inside a numbered-options
+  block (`1. Run / 2. Cancel`) and wait. `1` proceeds; anything else aborts.
 
 ### 4. Run the CLI
 
-Once the user picks `1`, invoke the same arguments with `run` plus
+Invoke the same arguments with `run` plus
 `--confirm` and an output path under `agents/runtime/council/sessions/`:
 
 ```bash

package/dist/agent-src/commands/council.md

@@ -35,7 +35,7 @@ sub-command swaps the mode-specific addendum.
 
 1. Resolve the target + capture the original ask.
 2. Check the council is configured + price table fresh.
-3. Cost confirmation (ALWAYS ASK for billable members).
+3. Spend bound (ask only when a billable member has no ceiling — ADR-230).
 4. Run the CLI.
 5. Render the report (5 / 5a / 5b — render → critical lens → user options).
 6. Hard floor — text only.

package/dist/agent-src/commands/design-system/generate.md

@@ -26,13 +26,35 @@ recommendation. Args: `"<product / industry / keywords>"`.
 ```bash
 npx tsx node_modules/@event4u/agent-config/src/skills/corpus-grounding/scripts/ground.ts \
     ground "<query>" \
-    --manifest node_modules/@event4u/agent-config/src/skills/design-intelligence/data/manifest.json
+    --manifest node_modules/@event4u/agent-config/src/skills/design-intelligence/data/manifest.json \
+    [--variance 1-10] [--motion 1-10] [--density 1-10]
 ```
 
 Add `--persist <dir> --project-name <name>` to write
 `design-system/<slug>/MASTER.md`, and `--page <name>` for a page-override stub
 alongside it. The engine is read-only without `--persist`.
 
+## The three dials — optional, and unset means unset
+
+Each takes an integer 1-10. A dial you do not pass changes nothing at all, so
+the no-flag result is exactly what it was before dials existed.
+
+| Dial | 1-3 | 4-7 | 8-10 | What it touches |
+|---|---|---|---|---|
+| `--variance` | Centered / Minimal | Balanced / Modern | Bold / Asymmetric | Biases which style is **selected** from the rows the corpus returned |
+| `--motion` | Subtle | Standard | Complex | Pulls a motion recipe at that intensity tier |
+| `--density` | Spacious | Standard | Dense / Dashboard | Overrides the spacing scale |
+
+**`--variance` biases selection, never retrieval.** The rows that come back from
+the corpus are the same with or without it; the dial only changes which of those
+rows is picked. And when it changes the pick away from what the grounding rule's
+own keywords would have chosen, the run **says so in its evidence gaps** — a
+preference is allowed to win, but never quietly.
+
+`--motion` needs the `gsap` domain (`motion.csv`) in the manifest. Without it,
+or when no row matches the requested tier, the run reports that rather than
+returning an empty block.
+
 ## After the recommendation
 
 Offer both destinations, and let the human pick — they are different artifacts,

package/dist/agent-src/commands/pr/create.md

@@ -68,52 +68,77 @@ gate.
 
 Run, in order:
 
-1. `git fetch origin {target-base} --quiet` (default: `main`).
-2. `git rev-list --count HEAD..origin/{target-base}` — number of
-   commits HEAD is **behind** the base.
-3. `gh pr list --state open --base {target-base} --limit 20 --json number,headRefName,files`
-   — open PRs targeting the same base.
+1. **Ask the gate, never hand-roll the comparison:**
+
+   ```bash
+   ./scripts-run src/scripts/check_branch_freshness
+   ```
+
+   It resolves the base itself — explicit `--base`, else the base of the **open
+   PR for this branch** as the forge reports it, else the repo default from
+   `origin/HEAD`. Do **not** substitute `git rev-list --count HEAD..origin/main`:
+   that hardcodes a base this branch may not target, and it reads the local
+   tracking ref, which is a fetch from earlier in the session — a memory, not a
+   check (`direct-answers` Iron Law 2). Exit `0` = current, `1` = behind, and the
+   refusal names the base **and where that base came from**.
+
+2. `gh pr list --state open --base {resolved-base} --limit 20 --json number,headRefName,files`
+   — open PRs targeting the same base, for the overlap question below.
 
 **Decision matrix:**
 
-| Behind | Overlapping open PR touches same files? | Action |
+| Gate | Overlapping open PR touches same files? | Action |
 |---|---|---|
-| `0` | — | Proceed to Step 2 |
-| `1–N` | No | **Merge `origin/main` in** — `git merge origin/{target-base} --no-edit` — then proceed. No need to ask; state that you did it. Abort + surface only on conflict. |
-| `1–N` | **Yes** | STOP — surface the overlapping PR number, ask: stack on top of it / wait for it to land / proceed-anyway-and-accept-conflicts / cancel |
+| exit `0` | — | Proceed to Step 2 |
+| exit `1` | No | **Merge the base in** — `git fetch origin && git merge origin/{resolved-base} --no-edit` — then run the regeneration set below, then proceed. No need to ask; state that you did it. |
+| exit `1` | **Yes** | STOP — surface the overlapping PR number, ask: stack on top of it / wait for it to land / proceed-anyway-and-accept-conflicts / cancel |
+| warns `could not ask the forge` | — | The **default** base was checked and an open PR against a different base was **not** ruled out. Say so; do not report it as a clean freshness pass. |
 
-A branch behind its base is **not** current — bring `main` in **before** the PR
+A branch behind its base is **not** current — bring the base in **before** the PR
 exists, never after. Auto-merge when there is no file overlap (the common case);
 only ask when an overlapping open PR makes the base genuinely ambiguous. Never
 improvise the base or silently proceed when behind **and** overlapping. The
 10-second fetch beats hours of rebase reconciliation after the parent PR lands.
 
-**On merge conflict: regenerate, don't abort.** Most conflicts on this class of
-branch are in **generated files** — the roadmap dashboard
-(`agents/roadmaps-progress.md`), `docs/proof.md`, the `dist/agent-src/`
-projection, tool projections. Those are resolved **by regenerating from the merged sources**,
-never by hand-picking hunks:
-
-1. `./agent-config roadmap:progress` → resolves the dashboard.
-2. `./scripts-run src/scripts/build_proof` → resolves `docs/proof.md`.
-3. `bash src/scripts/condense.sh --sync`, then `--check` → resolves any
-   conflict inside `dist/agent-src/` by re-projecting from the merged source.
+**After EVERY merge — not only on conflict — regenerate the derived files.**
+This is the half that gets skipped, because a merge that lands cleanly *looks*
+finished. It is not: these files are **generated**, so a clean auto-merge of
+them produces a file that describes neither side's sources. The dashboard is the
+usual casualty — it merges without a conflict marker and then reports a roadmap
+set that no longer exists.
+
+1. `./agent-config roadmap:progress` → the roadmap dashboard.
+2. `./scripts-run src/scripts/build_proof` → `docs/proof.md`.
+3. `bash src/scripts/condense.sh --sync`, then `--check` → re-projects
+   `dist/agent-src/` from the merged source.
 4. Stage the regenerated files, complete the merge, and re-run the touched
    verification before pushing.
 
-Only a conflict in **hand-authored** content (source code, prose you or the
-other PR wrote) is surfaced to the user; a generated-file conflict never is.
+On a **conflict**, the same four steps are the resolution — regenerate from the
+merged sources, never hand-pick hunks. Only a conflict in **hand-authored**
+content (source code, prose you or the other PR wrote) is surfaced to the user;
+a generated-file conflict never is.
 
 ### 1b-ii. Update freshness — MANDATORY on EVERY later push to an open PR
 
 The gate above is not creation-only. **Every subsequent push** to a branch with
 an open PR (a CI fix, a review response, a follow-up commit) re-runs the same
-sequence first: `git fetch origin {target-base}` → behind-check → merge
-`origin/{target-base}` in (regenerating on conflict as above) → verify → push.
+sequence first: `check_branch_freshness` → on exit `1`, merge
+`origin/{resolved-base}` in → regenerate the derived files → verify → push.
 A PR that sits open while its base advances goes stale silently; keeping the
 base merged **at every touch** means it stays `mergeStateStatus: CLEAN` instead
-of accumulating conflicts for the moment the user wants to merge. If the
-behind-check shows `0`, the step costs one fetch and nothing else.
+of accumulating conflicts for the moment the user wants to merge. If the gate
+exits `0`, the step costs one `ls-remote` and nothing else.
+
+Once the PR exists, this is also when the resolution gets *more* precise rather
+than less: the gate can now read the PR's real `baseRefName` from the forge, so
+a stacked or release-line PR is measured against the branch it will actually
+merge into instead of against the repo default.
+
+Where the project wires this gate into its pre-push chain, a push that goes out
+through the hook is covered without a separate step. Run it by hand when you push
+with the hook bypassed, or when you are about to open the PR some time after the
+last push — that gap is exactly where a base moves unseen.
 
 ### 1c. PR-gate — archive completed roadmaps (MANDATORY)
 

package/dist/agent-src/commands/worktree/create.md

@@ -26,7 +26,7 @@ Thin orchestration — mechanics live in the skills; do not restate them.
 ### 1. Gate + create
 
 Run [`using-git-worktrees`](../../../skills/using-git-worktrees/SKILL.md)
-in full: `worktrees.mode` pre-flight, existing-worktree inspection,
+in full: instruction-only pre-flight, existing-worktree inspection,
 directory convention, ignore-safety check, worktree creation, the
 [§ 4b seeding allow/deny list](../../../skills/using-git-worktrees/SKILL.md#4b-seed-the-worktree--allow--deny-list),
 dependency install + clean baseline. Its Iron Law (no worktree without
@@ -61,5 +61,7 @@ name it as the suggested next action — suggest, never auto-run.
 ### Rules
 
 - **Do NOT commit or push.**
-- Worktree creation stays permission-gated per `scope-control` unless
-  `worktrees.mode: on` grants standing permission (skill § 0).
+- Worktree creation is instruction-only (ADR-229). Invoking this command
+  **is** the explicit request, so the skill's § 0 pre-flight passes and no
+  further permission question is raised. Nothing here creates a worktree
+  the user did not ask for.

package/dist/agent-src/rules/settings-ask-protocol.md

@@ -77,12 +77,16 @@ conservative default (`""`, `false`, `false`), so absent is indistinguishable
 from *no* and never from *yes*.
 
 **Class C carrying an `ask` value in its own enum** — a runtime question whose
-answer dies with the run: `worktrees.mode` (**ships** as `ask`), plus
-`tokens.rich_skills`, `subagents.adversarial_council`, `decision_engine.on_block`
-(which can be set to `ask` by a human but do not ship that way). Always-on
-orchestration (road-to-always-on-orchestration Phase 1) deleted
-`subagents.auto` and `subagents.budget_routing` — the layer no longer carries a
-setting for this protocol to route.
+answer dies with the run: `tokens.rich_skills`,
+`subagents.adversarial_council`, `decision_engine.on_block`. **None of them
+ships as `ask`** — a human sets it, so the whole class is opt-in and no shipped
+default routes a question through this protocol at all.
+
+Two deletions got the list here. Always-on orchestration
+(road-to-always-on-orchestration Phase 1) removed `subagents.auto` and
+`subagents.budget_routing`. ADR-229 removed `worktrees.mode`, which was the last
+key that *shipped* as `ask` — worktree creation is instruction-only now, so
+there is no setting and nothing to ask about.
 
 ## The budget
 
@@ -153,9 +157,10 @@ by `settings:set` and only *displayed* by the GUI.
 - Class A — resolves to its default, never asked.
 - Class C **not** set to `ask` — no question; a change is a human edit.
 - Already decided on any resolution layer, or the run completes without it.
-- **The `ask` routes to a permission gate.** `worktrees.mode: ask` means
-  `scope-control`'s per-creation gate applies — a permission ask about an
-  *action*, which that rule owns; a storage line there would store nothing.
+- **The `ask` routes to a permission gate** — a permission ask about an
+  *action*, which `scope-control` owns; a storage line there would store
+  nothing. `worktrees.mode: ask` was the worked example until ADR-229 deleted
+  the key; the carve-out is kept because the shape recurs, not the key.
 - **The `ask` is code, not agent-carried.** `decision_engine.on_block: ask` is a
   TTY prompt in `work_engine/hooks/builtin/decision_gate.ts`.
 

package/dist/agent-src/skills/ai-council/SKILL.md

@@ -328,9 +328,25 @@ not a default.
 2. **Bundle + redact.** Call `scripts/ai_council/bundler.ts` to produce
    a redacted artefact bundle. If `BundleTooLarge` fires, surface the
    size and ask the user to narrow scope — do NOT truncate silently.
-3. **Confirm spend.** Before any network call, surface members + cost
-   ceiling and require an explicit user `1` to proceed. Autonomy
-   settings do not override this gate.
+3. **Clear the spend bound.** The council is standing-authorized — there is
+   no per-invocation approval. What still binds is a *bound*, resolved in
+   this order:
+   - **No billable member** (every member is `mode: cli` under
+     subscription auth — `billable=False`) → nothing is spent. Render the
+     estimate table as information and fan out. No gate.
+   - **Billable members with a configured ceiling** (`cost_budget.max_total_usd`
+     non-zero, or `daily_limit_usd` non-zero) → the ceiling **is** the
+     authorization the user already gave. Render the estimate and fan out.
+     `on_overrun` still fires per member on breach (below) — that is where
+     the user regains the decision.
+   - **Billable members with no ceiling at all** (`max_total_usd: 0` AND
+     `daily_limit_usd: 0`) → unbounded paid spend, so surface the estimate
+     and require an explicit `1`. Nothing bounds the call, so the user must.
+
+   Autonomy settings do not *create* a bound and do not lift the last case.
+   Consumers who want the old per-run gate back set a small
+   `cost_budget.max_total_usd`: every call then breaches and `on_overrun`
+   asks, per member.
 4. **Fan out.** Dispatch the bundle to each enabled council member via
    `scripts/ai_council/orchestrator.ts`. Each member receives the
    neutrality preamble from `prompts.ts` plus the artefact — nothing
@@ -594,8 +610,9 @@ overrides the table regardless of how the lens resolved.
   identity — that primes the reviewer and collapses diversity.
 - Do NOT silently truncate a too-large bundle — surface the size and
   ask for narrower scope.
-- Do NOT auto-spend tokens under `personal.autonomy: on` — the cost
-  gate fires every time, no exceptions.
+- Do NOT spend against an unbounded budget without asking — a billable
+  member with neither `max_total_usd` nor `daily_limit_usd` set has no
+  ceiling, and autonomy is not a ceiling.
 - Do NOT reuse SDK clients across invocations — re-load keys via
   `load_*_key()` each call.
 
@@ -620,7 +637,8 @@ Real failure modes seen in the wild:
 | Paste the host-agent identity ("I am Augment / Claude Code…") | Identity primes the reviewer's model. | Neutrality preamble in `prompts.ts` already handles this. |
 | Silently truncate a too-large bundle. | Misleads the reviewer into thinking they saw the whole thing. | Bundler raises `BundleTooLarge`; surface and ask for narrower scope. |
 | Reuse the same SDK client across calls without re-loading the key. | Leaks the key in long-lived process state. | Each invocation builds fresh clients from `load_*_key()`. |
-| Auto-spend tokens under `personal.autonomy: on`. | Autonomy ≠ permission to spend money. | Always ask before consultation, even under autonomy. |
+| Spend against a billable member with no configured ceiling because "the council is standing-authorized". | Standing authorization is a *bound* the user set once, not a blank cheque. No ceiling = nothing was authorized. | Surface the estimate and ask, exactly in that one case (Procedure § 3). |
+| Re-ask per invocation once a ceiling exists. | The ceiling already carried the decision; re-asking is the approval burden this default removed. | Fan out; let `on_overrun` ask on breach. |
 | Forward council convergence to the user as numbered options without a host verdict. | Convergence ≠ correctness; the council never saw the codebase. | Apply the *Critical evaluation* lens; tag every finding `accept` / `accept-with-modification` / `reject` / `needs-input` with one-line reason. |
 | Reject a finding on preference, not evidence. | "I don't like this" is not a verdict. | Cite the file, line, ADR, or contract that justifies the rejection — or surface as `needs-input`. |
 | Paraphrase council output into the host's own analysis to defend a verdict. | Strips attribution, breaches `direct-answers` no-invented-facts. | Verdict cites host evidence (file:line); council output stays attributed in the per-member sections. |

package/dist/agent-src/skills/corpus-grounding/SKILL.md

@@ -71,7 +71,8 @@ Scripts resolve all paths **skill-relative** (per
 
 ./scripts-run <skills-root>/corpus-grounding/scripts/ground ground \
   --manifest <skills-root>/<domain-skill>/data/manifest.json \
-  "luxury e-commerce" [--context '{"data_heavy": true}'] [--persist DIR]
+  "luxury e-commerce" [--context '{"data_heavy": true}'] [--persist DIR] \
+  [--variance 1-10] [--motion 1-10] [--density 1-10]
 
 ./scripts-run <skills-root>/corpus-grounding/scripts/ground validate --manifest …
 ```
@@ -79,6 +80,16 @@ Scripts resolve all paths **skill-relative** (per
 `<skills-root>` is wherever skills are deployed (`~/.claude/skills/` for
 Claude Code installs; `src/skills/` inside this repo).
 
+**The three `ground` dials are optional 1-10 preferences, and unset means
+unset** — a dial you do not pass changes nothing, so the no-flag result is
+byte-identical to what it was before they existed. `--variance` biases which
+style is **selected** from the rows the corpus returned (never which rows come
+back), and when it changes the pick away from the grounding rule's own choice
+the run says so in its evidence gaps. `--motion` needs a `gsap` domain in the
+manifest and reports when a tier matches nothing. `--density` overrides the
+spacing scale. All three render in the text output and in a persisted
+`MASTER.md`, not only under `--json`.
+
 ## Output format
 
 **Interface contract v1 — stability promise.** Per ADR-061 §2 ("Opus