docs(skill): keep access-key creation out of agent sessions

skills/ocx/SKILL.md

@@ -90,6 +90,13 @@ starring would be useful, say so and let the user decide.
 The same boundary covers the session-gated `/api/codex-prompt` writes: read them with
 `ocx inspect codex-prompt`, and leave the writes to the dashboard.
 
+## Secret-bearing commands
+
+**Do not run `ocx access key create`.** Its one-time response contains the new plaintext
+data-plane credential, and command output can enter the agent transcript. Ask the user to create
+the key themselves in a terminal outside the agent session and confirm only the new key id before
+continuing with inspection or removal commands. Never ask the user to paste the key into chat.
+
 ## Destructive verbs
 
 `storage trash restore` and `storage policy run` refuse without `--yes` (exit 2, nothing sent).

skills/ocx/references/03_recipes.md

@@ -93,20 +93,27 @@ Read `accounts[]`. Two things to respect:
 `providers[]` and `models[]` carry `estimatedCostUsd`. Costs are estimates; `estimateReasons` in the
 log rows tells you why (for example `usage_estimated`, `expected_price_overlay`).
 
-## 5. Rotate an access key and confirm it went quiet
+## 5. Prepare an access-key rotation without exposing the new key
 
 ```bash
 ocx access key list --json
-ocx access key create rotated --json          # the plaintext key is in THIS response only
+```
+
+Creating a key returns its one-time plaintext credential. **Do not run the create command from an
+agent session:** terminal output can enter the model transcript. Ask the user to create the key
+themselves in a separate terminal and report only the new key id, never the key itself.
+
+After the user confirms that the replacement is configured, remove the old key:
+
+```bash
 ocx access key remove <old-id> --yes --json
 ocx access key list --json                    # the old id is gone; check usage on the rest
 ```
 
-Note the argument style: `create <name>` and `remove <id>` are **positionals**, not `--label` and
-`--id`. `remove` also refuses without `--yes`.
+Note the argument style: `remove <id>` is positional, not `--id`, and refuses without `--yes`.
 
 The list carries per-key usage, so a key whose count stops advancing is genuinely unused. The
-plaintext key appears once, in the `create` response, and is never retrievable again.
+replacement key's plaintext is never retrievable after creation.
 
 An `ambiguous` footer on the list means two configured keys share an id, so per-key totals do not
 exist for them — do not attribute usage to either.

tests/skill-ocx.test.ts

@@ -146,6 +146,14 @@ describe("the consent boundary is stated, not implied", () => {
     }
   });
 
+  test("agent-facing instructions never invoke the plaintext access-key creation command", () => {
+    const skill = readFileSync(SKILL, "utf8");
+    const recipes = read("references/03_recipes.md");
+    expect(skill).toContain("Do not run `ocx access key create`");
+    expect(recipes).not.toMatch(/^\s*ocx access key create\b/m);
+    expect(recipes).toContain("never the key itself");
+  });
+
   test("destructive verbs are documented as requiring --yes", () => {
     const skill = readFileSync(SKILL, "utf8");
     const recipes = read("references/03_recipes.md");