diff --git a/.changeset/16927-agent-tools-retirement-citation.md b/.changeset/16927-agent-tools-retirement-citation.md
new file mode 100644
index 0000000000..820b091504
--- /dev/null
+++ b/.changeset/16927-agent-tools-retirement-citation.md
@@ -0,0 +1,24 @@
+---
+'@objectstack/spec': patch
+---
+
+The `agent.tools` rejection now says why ADR-0064 binds, so its `Proposed` status does not read as "not yet in force"
+
+An author who writes the retired `agent.tools` key gets the tombstone's
+prescription, which rests the rule on **ADR-0064** (*"an agent's tool set is the
+union of its surface-compatible skills' tools"*). Following that citation lands
+on a record whose own header reads `**Status**: Proposed (2026-06-22)` and
+carries a `🔶 Cloud-owned — superseded in part by cloud ADR-0025` callout. From
+the record itself an author cannot tell that the rule still binds them — the
+weaker reading is the one the metadata invites.
+
+ADR-0064 stays the cited authority, because it is the record that states the
+invariant the key violated; **ADR-0109** (`Accepted — implemented (Phase 1)`)
+names `agent.tools` nowhere and only *builds on* that invariant, so retargeting
+the citation would send the author to a record that does not contain the rule
+they broke. The message instead gains one clarifying clause: the `Proposed` /
+cloud-owned status scopes the **runtime** half (tool resolution, which lives in
+cloud `service-ai`), while the **authoring** half is in force in this repo and
+ADR-0109 is the in-repo record carrying it.
+
+Prose only — the rejection, the retirement and the accept set are unchanged.
diff --git a/content/docs/references/ai/agent.mdx b/content/docs/references/ai/agent.mdx
index 880bc15536..159db8c9f8 100644
--- a/content/docs/references/ai/agent.mdx
+++ b/content/docs/references/ai/agent.mdx
@@ -51,7 +51,7 @@ const result = AIModelConfigSchema.parse(data);
 | **lifecycle** | `{ id: string; description?: string; contextSchema?: Record<string, any>; initial: string; … }` | optional | [EXPERIMENTAL — not enforced] State machine defining the agent conversation flow and constraints. Parsed but no runtime consumer yet. |
 | **surface** | `Enum<'ask' \| 'build'>` | optional (default: `"ask"`) | Product surface this agent binds ('ask' \| 'build') — ADR-0063 §1 |
 | **skills** | `string[]` | optional | Skill names to attach (Agent→Skill→Tool architecture) |
-| **tools** | `never` | optional | [REMOVED] `agent.tools` was removed in @objectstack/spec 17 — use `skills`. An agent reaches exactly the tools its surface-compatible skills declare (ADR-0064), so move each reference into a skill: a platform tool by its registered name, or `action_<name>` for one of your own AI-exposed Actions. This is NOT a rename — there is no key the value moves to: the migration DELETES the key and emits a notice naming each tool that was listed, and you re-declare each one in a skill by hand. Run `os migrate meta --from 16` to list the mechanical edits for existing sources; apply them by hand. |
+| **tools** | `never` | optional | [REMOVED] `agent.tools` was removed in @objectstack/spec 17 — use `skills`. An agent reaches exactly the tools its surface-compatible skills declare (ADR-0064), so move each reference into a skill: a platform tool by its registered name, or `action_<name>` for one of your own AI-exposed Actions. This is NOT a rename — there is no key the value moves to: the migration DELETES the key and emits a notice naming each tool that was listed, and you re-declare each one in a skill by hand. ADR-0064 itself still reads `Proposed` and is cloud-owned — that scopes its RUNTIME half (tool resolution, which lives in cloud `service-ai`), not this rejection: the authoring invariant binds you here, and ADR-0109 (Accepted — implemented) is the in-repo record that carries it. Run `os migrate meta --from 16` to list the mechanical edits for existing sources; apply them by hand. |
 | **knowledge** | `never` | optional | [REMOVED] `agent.knowledge` was removed in @objectstack/spec 17.0.0 (audit close-out) — declaring knowledge sources/indexes on an agent never scoped retrieval: the `search_knowledge` tool takes `sourceIds` from the LLM's tool-call arguments, not from the agent record. Delete the block. Restrict retrieval at the knowledge-service / source level (per-source permissions), and describe intended grounding in `instructions` so the model asks for the right sources. Run `os migrate meta --from 16` to list the mechanical edits for existing sources; apply them by hand. |
 | **active** | `boolean` | optional (default: `true`) |  |
 | **access** | `string[]` | optional | Who can chat with this agent |
diff --git a/packages/spec/src/ai/agent.zod.ts b/packages/spec/src/ai/agent.zod.ts
index 690dc4e58e..db3e7e4e16 100644
--- a/packages/spec/src/ai/agent.zod.ts
+++ b/packages/spec/src/ai/agent.zod.ts
@@ -239,6 +239,10 @@ export const AgentSchema = lazySchema(() => strictObject({
     'This is NOT a rename — there is no key the value moves to: the migration ' +
     'DELETES the key and emits a notice naming each tool that was listed, and ' +
     'you re-declare each one in a skill by hand. ' +
+    'ADR-0064 itself still reads `Proposed` and is cloud-owned — that scopes its ' +
+    'RUNTIME half (tool resolution, which lives in cloud `service-ai`), not this ' +
+    'rejection: the authoring invariant binds you here, and ADR-0109 ' +
+    '(Accepted — implemented) is the in-repo record that carries it. ' +
     'Run `os migrate meta --from 16` to list the mechanical edits for existing sources; apply them by hand.',
   ),