feat: unify agent actions and give Codex live control

AGENTS.md

@@ -46,18 +46,21 @@ Run what you build before calling it done. **If you can't run it, say so.**
 ## PR discipline
 
 - Paste the full PR URL, not the number.
+- Record it through the declared surface too: prose URLs are inert, so use
+  `@agency:pr.link(url: "<full URL>", item: "<item id>")` or the `pr:` field on
+  `items.state` when shipping an item.
 - When the change lands on master, close the PR and delete the branch in the same breath
   (`gh pr merge --delete-branch`). An open PR whose content already shipped reads as
   unfinished work.
 
 ## Closing an item you were given
 
-Reply checkboxes edit the project's item list: `- [ ]` proposes, `- [x] <exact title>`
-closes, `- [-] <exact title>` strikes an obsolete row. Titles match exactly and
-case-insensitively, so a paraphrase silently appends a near-duplicate. Full contract:
-[`docs/task-manager.md`](docs/task-manager.md#the-project-session-contract-three-checkboxes).
-Read that, not `projects.rs`: the verbs arrived late, and an older tree reads as
-append-only. [`wt-tools`](crates/wt-tools) reads the list and never writes.
+The project prompt supplies item ids and the declared Prompt Syntax surface. Report
+state with `@agency:items.state`, create with `@agency:items.add`, and remove an
+incorrect row with `@agency:items.retire`. Never address an existing row by title:
+paraphrases were how the old checkbox contract created near-duplicates. Full contract:
+[`docs/task-manager.md`](docs/task-manager.md#the-project-session-contract).
+[`wt-tools`](crates/wt-tools) reads the list and never writes.
 
 - **Close in the same turn the work ships.** Shipped means merged and released.
 - **Read the title before striking it.** `wt-tools search-items <word>` prints it as stored.

AgencyZero.md

@@ -3,21 +3,18 @@
 3. Ship through a pull request.
 4. Merge only after review passes, or the owner overrides.
 5. No em dashes.
-6. Track the work as items, in your reply, as it happens:
-   `- [ ]` opens, `- [~]` plans, `- [/]` starts, `- [>] title (#35)` ships.
-   `- [?]` when you are stuck on an answer only the owner can give.
-   Never `- [x]`: you cannot confirm your own fix works. The owner closes it.
-   Quoted or fenced checklists are never read as items, so quote freely.
-7. To retire a row, or to write into another project, ask the host directly:
-   `<ps @agency:items.inject(project: "ui")>` … checkbox lines … `</ps>`.
-   Project by name or by id, omit it to mean this one. Prose can open, plan,
-   start and ship a row here; only a directive can delete or reach elsewhere.
-8. Report item state by id, on its own line, as it happens:
+6. Track work through the declared Prompt Syntax surface, in your reply, as it
+   happens. Prose, checkboxes, quotes, fences and URLs are inert.
+7. Report item state by id, on its own line:
    `<ps @agency:items.state(id: "item-a3f9", status: "active")>`
    `<ps @agency:items.add(ref: "t1", title: "One line", status: "planning")>`
    `<ps @agency:items.retire(id: "item-a3f9")>` removes one that should not
    be there. The turn's prompt lists the open items and their ids, and answers
    back with what each directive did. Never address a row by its title.
+8. A pull request is state only when authored as state:
+   `<ps @agency:pr.link(url: "https://github.com/owner/repo/pull/35", item: "item-a3f9")>`.
+   Also paste the full URL in prose so the owner can open it. A run started from
+   an item ships with `items.state(..., status: "shipped", pr: "<full URL>")`.
 9. Ask before installing anything: a download, a global cache, a browser, a
    toolchain. A repo doc recommending it is not permission.
 10. Persisted data goes in a WorkTable table. Not a JSON file beside the store,

Cargo.toml

@@ -11,7 +11,7 @@ members = [
 ]
 
 [workspace.package]
-version = "0.1.55"
+version = "0.1.56"
 edition = "2024"
 publish = false
 

apps/gui/Cargo.toml

@@ -11,7 +11,7 @@ tauri-build = { version = "2", features = [] }
 [dependencies]
 # Carries the store forward when a column changes; see crates/wt-migrate.
 wt-migrate = { path = "../../crates/wt-migrate" }
-agent-abstraction = "0.4.4"
+agent-abstraction = "0.4.5"
 az-core.workspace = true
 worktable = "0.9"
 # The `worktable!` macro emits code that names these by bare path rather than

apps/gui/frontend/src/api/fixtures.ts

@@ -412,8 +412,8 @@ export const AGENT_STATUS: AgentStatus[] = [
       events: true,
       nativeSystem: false,
       commands: false,
-      liveFollowUp: false,
-      approvals: false,
+      liveFollowUp: true,
+      approvals: true,
     },
     checkedAt: ago(2 * 60_000),
   },

apps/gui/frontend/src/features/settings/SettingsTab.tsx

@@ -302,10 +302,7 @@ export function SettingsTab(): JSX.Element {
                     onChange={(defaultEffort) => void actions.saveSettings({ defaultEffort })}
                   />
                 </Row>
-                <Row
-                  label="Completed items"
-                  hint="what a session's own “done” report does to the row"
-                >
+                <Row label="Completed items" hint="what marking an item finished does to the row">
                   <PillMenu
                     label="Completed items"
                     value={current().completedItems}

apps/gui/frontend/src/stores/models.test.tsx

@@ -203,14 +203,14 @@ describe("what the prompt offers", () => {
     expect(sent?.model).toBe("gpt-5.6-sol");
   });
 
-  it("moves Ask to read-only when switching to OpenAI", async () => {
+  it("keeps Ask when switching to an approval-capable OpenAI model", async () => {
     const workspace = await mountWorkspace();
     workspace.actions.setTabModel("worktable", "claude", "sonnet", "ask");
     workspace.actions.setTabModel("worktable", "codex", "gpt-5.6-sol", "ask");
 
     const tab = workspace.state.tabs.find((candidate) => candidate.key === "worktable");
     expect(tab?.agent).toBe("codex");
-    expect(tab?.permission).toBe("read_only");
+    expect(tab?.permission).toBe("ask");
   });
 
   it("restores the last provider when a project is closed and reopened", async () => {
@@ -322,15 +322,15 @@ describe("posture follows Settings too", () => {
     expect(workspace.activeTab().permission).toBe("read_only");
   });
 
-  it("opens a Codex draft as read-only when Settings says Ask", async () => {
+  it("opens a Codex draft with Ask when Settings says Ask", async () => {
     const workspace = await mountWorkspace();
     await workspace.actions.saveSettings({ defaultAgent: "codex", defaultPermission: "ask" });
 
     workspace.actions.openDraft();
 
     await waitFor(() => expect(workspace.activeTab().kind).toBe("draft"));
     expect(workspace.activeTab().agent).toBe("codex");
-    expect(workspace.activeTab().permission).toBe("read_only");
+    expect(workspace.activeTab().permission).toBe("ask");
   });
 
   it("keeps Copilot out of project tabs until project runs support it", async () => {

apps/gui/frontend/src/stores/workspace.test.tsx

@@ -66,7 +66,7 @@ describe("startup", () => {
   it("derives approval postures from provider capabilities", async () => {
     const workspace = await mountWorkspace();
     expect(workspace.permissionsFor("claude")).toContain("ask");
-    expect(workspace.permissionsFor("codex")).not.toContain("ask");
+    expect(workspace.permissionsFor("codex")).toContain("ask");
   });
 });
 

apps/gui/frontend/src/types/index.ts

@@ -477,8 +477,8 @@ export interface GlobalSettings {
   forwardProxyVars: boolean;
   notifications: NotificationSettings;
   /**
-   * What a session's `- [x]` does to an existing item: `"resolve"` marks it
-   * finished and keeps the row, `"delete"` removes it outright.
+   * What the owner's finished action does to an existing item: `"resolve"`
+   * keeps the finished row, while `"delete"` removes it outright.
    */
   completedItems: "resolve" | "delete";
   /** How the workspace is coloured. See {@link ThemeSettings}. */

apps/gui/src/agents.rs

@@ -249,8 +249,8 @@ mod tests {
         assert!(claude.live_follow_up);
         assert!(claude.approvals);
         assert!(claude.commands);
-        assert!(!codex.live_follow_up);
-        assert!(!codex.approvals);
+        assert!(codex.live_follow_up);
+        assert!(codex.approvals);
         assert!(!codex.commands);
     }
 }

apps/gui/src/db/schema/pull_request.rs

@@ -1,6 +1,6 @@
 //! Pull requests cut during a project's runs, tracked as chips over the composer.
 //!
-//! One row per PR URL per project, upserted when a reply mentions the URL and
+//! One row per PR URL per project, created from an authored PS directive and
 //! refreshed through `gh` when available. `state` is GitHub's own word
 //! (`OPEN` | `MERGED` | `CLOSED`, or `unknown` before the first refresh);
 //! `ci` is the check rollup reduced to `pass` | `fail` | `pending` | `none` |