docs(liveness): stop counting authoring/preview renderers as runtime consumers (#1878) (#3685)

A 2026-06 sweep that "missed objectui" over-corrected and marked 13 properties
`live` on the evidence of a metadata-admin previews/*Preview.tsx panel — a
surface that only echoes what the author typed. `live` is supposed to mean
"authoring this changes runtime behaviour".

Three of the 13 were re-verified in the §3 naming-drift recheck; all three
were wrong or misleading:

- skill.permissions -> dead + authorWarn. The cloud SkillRegistry reads only
  active/triggerConditions/tools; nothing gates invocation by this list.
  Mirrors the identical tool.permissions entry. (Cloud documents the intent
  in a comment without implementing it.)
- agent.knowledge -> dead + authorWarn. search_knowledge takes sourceIds from
  the LLM's tool-call arguments, never from the agent record, so a declared
  knowledge scope constrains nothing (and an agent can search sources it was
  never scoped to).
- action.disabled: verdict kept (it IS live now) but evidence corrected to the
  six rendering surfaces. At ledger-writing time it was enforced on ONE of
  six, so the preview-based verdict hid a real five-surface silent no-op —
  fixed in objectui#2863.

Also: sync the hand-maintained count table (agent 14/5/0 -> 13/5/1, skill
10/-/- -> 9/-/1) and add a "an authoring/preview renderer is NOT a runtime
consumer" section naming the ten still-unverified preview-only claims, with
the tie-break rule: when in doubt prefer dead + authorWarn — a warning on a
property that works costs nothing; silence on a property that does nothing
ships a bug.

check:liveness green (exit 0); the 13 unregistered-proof warnings are
pre-existing on main.

Co-authored-by: Jack Zhuang <277994282+os-zhuang@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

Author: os-zhuang

.changeset/liveness-ledger-ai-scope-honesty.md

@@ -0,0 +1,12 @@
+---
+---
+
+docs(liveness): correct three ledger entries that counted an authoring/preview renderer as a runtime consumer (#1878 §3 recheck)
+
+`skill.permissions` and `agent.knowledge` were marked `live` citing only a
+metadata-admin PREVIEW panel — which echoes what the author typed and proves
+nothing about enforcement. Both are corrected to `dead` + `authorWarn` with an
+actionable hint. `action.disabled`'s evidence is corrected to the six real
+rendering surfaces (its verdict was right for the wrong reason and hid a
+five-surface silent no-op, since fixed in objectui#2863). Ledger-only; releases
+nothing.

packages/spec/liveness/README.md

@@ -46,6 +46,29 @@ Resolution per property: **ledger entry → spec `.describe()` marker → UNCLAS
 Framework provenance/lock fields (`_lock*`, `_provenance`, `_packageId/Version`,
 `protection` — ADR-0010) are auto-classified `live`.
 
+### ⚠️ An authoring/preview renderer is NOT a runtime consumer
+
+`live` means **authoring the property changes runtime behaviour**. A Studio
+`*.form.ts` input or a `metadata-admin/previews/*Preview.tsx` panel merely
+*echoes back what the author typed* — it proves the property round-trips, never
+that anything acts on it. A 2026-06 sweep that "missed objectui" over-corrected
+and marked **13 properties `live` citing only a preview renderer**; of those,
+three were re-verified in 2026-07 and **all three were wrong or misleading**:
+
+- `skill.permissions` — no gate anywhere (corrected → `dead` + `authorWarn`)
+- `agent.knowledge` — retrieval scope comes from the LLM's tool args (corrected → `dead`)
+- `action.disabled` — enforced on **one of six** rendering surfaces at the time;
+  the verdict was right for the wrong reason and hid a five-surface silent
+  no-op (evidence corrected; the gap itself fixed in objectui#2863)
+
+The remaining ten preview-only claims (`action.execute`/`shortcut`/`bulkEnabled`,
+`flow.status`/`active`, `skill.triggerPhrases`, `tool.category`/
+`requiresConfirmation`/`active`/`builtIn`) are **unverified** — treat them as
+suspect until someone cites a real runtime reader. When in doubt, the honest
+status is `dead` + `authorWarn`: an author who gets a warning for a property
+that turns out to work loses nothing; an author who gets silence for a property
+that does nothing ships a bug.
+
 ## Runtime proofs — prove-it-runs (ADR-0054)
 
 `live` today means only *a static pointer to a consumer* — proof that something
@@ -207,9 +230,9 @@ EOF
 | hook | 11 | – | 2 | – | model-healthy; only label/description dead (benign) |
 | permission | 32 | – | 0 | – | CRUD/FLS/RLS live; dead `contextVariables` REMOVED (ADR-0105 D11 — RLS resolves only the `current_user.*` built-ins plus runtime-staged `rlsMembership` sets) |
 | position | 4 | – | – | – | (role's ADR-0090 successor) fully live |
-| agent | 14 | 5 | 0 | – | dead `tenantId` + `planning.strategy`/`allowReplan` REMOVED (#2377) — only `planning.maxIterations` live; autonomy tier experimental |
+| agent | 13 | 5 | 1 | – | dead `tenantId` + `planning.strategy`/`allowReplan` REMOVED (#2377) — only `planning.maxIterations` live; autonomy tier experimental; `knowledge` CORRECTED to dead 2026-07 — `search_knowledge` takes `sourceIds` from the LLM's tool-call args, never from the agent record (#1878 §3 recheck) |
 | tool | 9 | 1 | 1 | – | `permissions` dead — tool invocation not permission-gated by it |
-| skill | 10 | – | – | – | fully live |
+| skill | 9 | – | 1 | – | `permissions` CORRECTED to dead 2026-07 — the cloud SkillRegistry reads only `active`/`triggerConditions`/`tools`; mirrors the identical `tool.permissions` entry (#1878 §3 recheck) |
 | dataset | 19 | – | 0 | – | `measures.certified` (declared-but-unenforced governance flag) REMOVED in 16.0 (#2377) |
 | page | 16 | – | – | 1 | fully live + one planned |
 | view | 70 | 0 | 5 | – | list/form drilled via `children` (#2998 Track B); dead = list.{responsive,performance} + form.{data,defaultSort,aria}, all but aria authorWarn'd; form.{buttons,defaults} now live — objectui ObjectForm folds them onto its flat props (framework#1894 / #2998); audit-era DEAD lines superseded by re-verification (submitBehavior, sharing.lockedBy, list ViewData providers, and the ADR-0021 chart shape — all live now); level-2 dead residue (userActions.buttons, addRecord.mode/formView, tabs[].order) noted on parents — one drill level only |

packages/spec/liveness/action.json

@@ -101,8 +101,8 @@
     },
     "disabled": {
       "status": "live",
-      "evidence": "objectui: packages/app-shell/src/views/metadata-admin/previews/ActionPreview.tsx:169",
-      "note": "LIVE via objectui metadata-admin authoring UI — the 2026-06 audit missed objectui; verified by reading the file."
+      "evidence": "objectui: components action-button/-group/-icon/-menu + app-shell DeclaredActionsBar + plugin-detail record-quick-actions — all six rendering surfaces gate the Button on the predicate",
+      "note": "EVIDENCE CORRECTED 2026-07 (#1878 §3 recheck): previously cited the metadata-admin PREVIEW renderer, which only echoes the authored value. At that time the predicate was actually enforced on ONE of six surfaces (#1885 wired action:button alone) — the 'live' verdict was right for the wrong reason and hid a real silent no-op on the other five. objectui#2863 wired the rest; showcase specimen `showcase_archive_task` (#3643) dogfoods it."
     },
     "shortcut": {
       "status": "live",

packages/spec/liveness/agent.json

@@ -65,9 +65,11 @@
       "note": "evaluateAgentAccess() — caller must hold ALL required permissions/roles; enforced at the chat route. #1884."
     },
     "knowledge": {
-      "status": "live",
-      "evidence": "objectui: packages/app-shell/src/views/metadata-admin/previews/AgentPreview.tsx:69 (d.knowledge; KnowledgeSummary renders sources/indexes)",
-      "note": "LIVE via objectui metadata-admin authoring UI — the 2026-06 audit missed objectui; verified by reading the file."
+      "status": "dead",
+      "evidence": "agent.form.ts + objectui AgentPreview.tsx:69 (both authoring-side echo only); the RAG path never reads it",
+      "authorWarn": true,
+      "authorHint": "Declaring `knowledge.sources` does NOT scope retrieval — the `search_knowledge` tool takes `sourceIds` from the LLM's tool-call arguments (cloud service-ai/src/tools/knowledge-tools.ts:96), never from the agent record. Restrict retrieval at the knowledge-service / source level instead.",
+      "note": "CORRECTED 2026-07 (#1878 §3 recheck): was marked live on the evidence of the metadata-admin PREVIEW renderer — an echo of what the author typed, not runtime consumption. The `topics`→`sources` NAMING drift is genuinely fixed (#1891 alias fold); the block is nonetheless behaviourally inert. Wiring it would also close a scope gap (an agent can search sources it was never scoped to) — tracked separately."
     },
     "lifecycle": {
       "status": "experimental",

packages/spec/liveness/skill.json

@@ -47,9 +47,11 @@
       "note": "LIVE via objectui metadata-admin authoring UI — the 2026-06 audit missed objectui; verified by reading the file."
     },
     "permissions": {
-      "status": "live",
-      "evidence": "objectui: packages/app-shell/src/views/metadata-admin/previews/SkillPreview.tsx:48",
-      "note": "LIVE via objectui metadata-admin authoring UI — the 2026-06 audit missed objectui; verified by reading the file."
+      "status": "dead",
+      "evidence": "skill.form.ts + objectui SkillPreview.tsx:48 (both authoring-side echo only); no gate in the cloud SkillRegistry",
+      "authorWarn": true,
+      "authorHint": "Skill invocation is NOT permission-gated by this list — the registry reads only `active`/`triggerConditions`/`tools`. Agent-level `access`/`permissions` ARE enforced (#1884), so gate at the agent, or gate the underlying tools' actions via permission sets (ADR-0066).",
+      "note": "CORRECTED 2026-07 (#1878 §3 recheck): was marked live on the evidence of the metadata-admin PREVIEW renderer — an echo of what the author typed, not runtime consumption. Mirrors the identical `tool.permissions` entry. Cloud documents the intent ('Layer permissions via Skill.permissions independent of the agent's permissions', service-ai-studio/src/skills/metadata-authoring-skill.ts:15) without implementing it — enforce-or-prune tracked separately."
     }
   }
 }