feat(skills): remove the skills a run installed when it ends

.claude/skills/adding-skill-program/SKILL.md

@@ -125,7 +125,8 @@ Outside the folder, the
 [TUI program registry](../../../src/tui/programs/index.ts) imports and spreads
 the entry. A program with no entry runs
 [AGENT_SKILL_STEPS](../../../src/tui/programs/shared/skill-flow.ts): intro,
-health check, auth, run, outro, and keep-skills, with the skill deck. Map
+health check, auth, run, and outro, with the skill deck. `runProgram`
+removes the skills a run installed when it ends. Map
 its steps to swap in your own intro screen, as
 [audit's flow](../../../src/tui/programs/audit/flow.ts) does. Auth also applies
 the shared [AI opt-in gate](../../../src/tui/ai-opt-in-gate.ts),

.claude/skills/exploring-the-wizard/SKILL.md

@@ -90,8 +90,8 @@ own decisions through the same state and action contract.
    dismissing it. `integration` says `failed` once the run's phase is `error`;
    a host exit can instead surface as a socket error.
 6. After successful agent completion, finish the offered outro and follow-up
-   actions. For the integration flow, `session.skillsComplete` marks the tail's
-   completion. Other programs can have a terminal outro or exit screen.
+   actions. `session.flowComplete` turns true once the flow reaches its exit
+   screen, which ends the run.
 
 | Decision                                 | Action and `params`                                                                          |
 | ---------------------------------------- | -------------------------------------------------------------------------------------------- |
@@ -103,10 +103,10 @@ own decisions through the same state and action contract.
 | Finish outro                             | `dismiss_outro`                                                                              |
 | Record MCP outcome                       | `set_mcp_outcome`, `{ outcome: "skipped" }` or `{ outcome: "installed", clients: [...] }`    |
 | Dismiss suggested prompts / Slack step   | `dismiss` / `dismiss_slack`                                                                  |
-| Record keep-skills choice                | `keep_skills`, `{ kept: true }` or `{ kept: false }`                                         |
 
-MCP and keep-skills actions commit store state; recording an outcome does not
-perform the corresponding installation or cleanup. Report which outcomes were
+The MCP action commits store state; recording an outcome does not perform the
+installation. `runProgram` removes the skills a run installed when the run
+ends, so the app has none left afterwards. Report which outcomes were
 simulated. A completed run also needs a review of the app diff and expected
 integration behavior before calling the integration valid.
 

README.md

@@ -719,8 +719,7 @@ Example prompt — explore against
 >    for any setup question, then `run_agent` at auth.
 > 4. Poll `read_state` until `integration` is `done` (or `failed` — then report
 >    `integrationError`), snapshotting as the run screen progresses.
-> 5. Finish the tail: `dismiss_outro`, `set_mcp_outcome`, then
->    `keep_skills`.
+> 5. Finish the tail: `dismiss_outro`, then `set_mcp_outcome`.
 >
 > Then show me the saved snapshots in order, the screen path, whether `posthog`
 > landed in the app, and anything that broke.

e2e-harness/ARCHITECTURE.md

@@ -106,10 +106,10 @@ again to refresh background status. Legal actions come from
 `read_state.actions`; the MCP has no `list_actions` or `wait_for_change` tool.
 
 Keep handling overlays while the agent runs. `runPhase=completed` ends the main
-work, while `session.skillsComplete` ends the integration flow's follow-up
-screens and the host with it. Recording MCP or keep-skills outcomes only commits
-store state; it does not perform installation or cleanup. For failures, capture
-the error outro before dismissing it: dismissing it ends the run.
+work, while `session.flowComplete` marks the flow reaching its exit screen,
+which ends the host. Recording an MCP outcome only commits store state; it does
+not perform the installation. For failures, capture the error outro before
+dismissing it: dismissing it ends the run.
 
 ## Current host limitations
 

e2e-harness/__tests__/__snapshots__/e2e-flow-snapshot.test.ts.snap

@@ -1,13 +1,12 @@
 // Vitest Snapshot v1, https://vitest.dev/guide/snapshot.html
 
-exports[`e2e flow snapshot — ai-observability > walks intro → health → auth → run → outro → skills 1`] = `
+exports[`e2e flow snapshot — ai-observability > walks intro → health → auth → run → outro → exit 1`] = `
 {
   "profile": {
     "ask": "first",
     "healthCheck": "dismiss",
     "mcp": "skip",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "ai-observability",
@@ -32,25 +31,17 @@ exports[`e2e flow snapshot — ai-observability > walks intro → health → aut
       "action": "dismiss_outro",
       "screen": "outro",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;
 
-exports[`e2e flow snapshot — metrics > walks intro → health → auth → run → outro → skills 1`] = `
+exports[`e2e flow snapshot — metrics > walks intro → health → auth → run → outro → exit 1`] = `
 {
   "profile": {
     "ask": "first",
     "healthCheck": "dismiss",
     "mcp": "skip",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "metrics",
@@ -75,13 +66,6 @@ exports[`e2e flow snapshot — metrics > walks intro → health → auth → run
       "action": "dismiss_outro",
       "screen": "outro",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;
@@ -94,7 +78,6 @@ exports[`e2e flow snapshot — posthog-integration > Next.js (with a setup quest
     "mcp": "skip",
     "notice": "keep",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "posthog-integration",
@@ -134,13 +117,6 @@ exports[`e2e flow snapshot — posthog-integration > Next.js (with a setup quest
       },
       "screen": "mcp",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;
@@ -176,13 +152,6 @@ exports[`e2e flow snapshot — posthog-integration > Node (no setup question) wa
       },
       "screen": "mcp",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;
@@ -234,7 +203,6 @@ exports[`e2e flow snapshot — self-driving > integration-first (no existing Pos
     "integrate": true,
     "mcp": "skip",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "self-driving",
@@ -293,7 +261,6 @@ exports[`e2e flow snapshot — upload-source-maps > walks a stable path 1`] = `
     "healthCheck": "dismiss",
     "mcp": "skip",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "error-tracking-upload-source-maps",
@@ -318,18 +285,11 @@ exports[`e2e flow snapshot — upload-source-maps > walks a stable path 1`] = `
       "action": "dismiss_outro",
       "screen": "source-maps-outro",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;
 
-exports[`e2e flow snapshot — warehouse-source > walks intro → auth → run → outro → skills 1`] = `
+exports[`e2e flow snapshot — warehouse-source > walks intro → auth → run → outro → exit 1`] = `
 {
   "profile": {
     "ask": "first",
@@ -383,7 +343,6 @@ exports[`e2e flow snapshot — warehouse-source > walks intro → auth → run 
     "mcp": "skip",
     "notice": "keep",
     "setup": "first",
-    "skills": "delete",
     "slack": "skip",
   },
   "program": "warehouse-source",
@@ -404,13 +363,6 @@ exports[`e2e flow snapshot — warehouse-source > walks intro → auth → run 
       "action": "dismiss_outro",
       "screen": "outro",
     },
-    {
-      "action": "keep_skills",
-      "params": {
-        "kept": false,
-      },
-      "screen": "keep-skills",
-    },
   ],
 }
 `;

e2e-harness/__tests__/__snapshots__/keyboard-equivalence.test.tsx.snap

@@ -4,12 +4,11 @@ exports[`keyboard commit vs control action commit > audit-outro: any key vs dism
 {
   "action": {
     "outroDismissed": true,
-    "screen": "keep-skills",
+    "screen": "exit",
   },
   "keyboard": {
     "outroDismissed": true,
-    "screen": "keep-skills",
-    "skillsComplete": true,
+    "screen": "exit",
   },
 }
 `;
@@ -29,15 +28,6 @@ exports[`keyboard commit vs control action commit > intro: enter on Continue vs
 }
 `;
 
-exports[`keyboard commit vs control action commit > keep-skills: mount with no skills dir vs keep_skills 1`] = `
-{
-  "action": {
-    "skillsComplete": true,
-  },
-  "keyboard": {},
-}
-`;
-
 exports[`keyboard commit vs control action commit > manual-auth-code: escape vs dismiss_auth_code 1`] = `
 {
   "action": {
@@ -73,7 +63,7 @@ exports[`keyboard commit vs control action commit > mcp: enter on install result
       "Claude Code",
     ],
     "mcpOutcome": "installed",
-    "screen": "keep-skills",
+    "screen": "exit",
   },
   "keyboard": {
     "mcpComplete": true,
@@ -82,8 +72,7 @@ exports[`keyboard commit vs control action commit > mcp: enter on install result
       "Claude Code",
     ],
     "mcpOutcome": "installed",
-    "screen": "keep-skills",
-    "skillsComplete": true,
+    "screen": "exit",
   },
 }
 `;
@@ -176,12 +165,11 @@ exports[`keyboard commit vs control action commit > source-maps-outro: any key v
 {
   "action": {
     "outroDismissed": true,
-    "screen": "keep-skills",
+    "screen": "exit",
   },
   "keyboard": {
     "outroDismissed": true,
-    "screen": "keep-skills",
-    "skillsComplete": true,
+    "screen": "exit",
   },
 }
 `;

e2e-harness/__tests__/e2e-flow-snapshot.test.ts

@@ -158,7 +158,7 @@ async function traceFlow(
       else store.setRunPhase(RunPhase.Completed);
     }
 
-    if (decision.done || store.skillsComplete) break;
+    if (decision.done || store.currentScreen === ScreenId.Exit) break;
   }
   return trace;
 }
@@ -229,7 +229,7 @@ describe('e2e flow snapshot — upload-source-maps', () => {
 describe('e2e flow snapshot — ai-observability', () => {
   const profile = profileFor(Program.AiObservability);
 
-  it('walks intro → health → auth → run → outro → skills', async () => {
+  it('walks intro → health → auth → run → outro → exit', async () => {
     expect({
       program: 'ai-observability',
       profile,
@@ -245,7 +245,7 @@ describe('e2e flow snapshot — ai-observability', () => {
 describe('e2e flow snapshot — metrics', () => {
   const profile = profileFor(Program.Metrics);
 
-  it('walks intro → health → auth → run → outro → skills', async () => {
+  it('walks intro → health → auth → run → outro → exit', async () => {
     expect({
       program: 'metrics',
       profile,
@@ -270,13 +270,13 @@ describe('e2e flow snapshot — metrics', () => {
       screen: 'metrics-intro',
       action: 'confirm_setup',
     });
-    expect(trace.at(-1)?.action).toBe('keep_skills');
+    expect(trace.at(-1)?.action).toBe('dismiss_outro');
     expect(trace.length).toBeLessThan(40);
   });
 });
 
 describe('e2e flow snapshot — warehouse-source', () => {
-  it('walks intro → auth → run → outro → skills', async () => {
+  it('walks intro → auth → run → outro → exit', async () => {
     expect({
       program: 'warehouse-source',
       profile: profileFor(Program.WarehouseSource),
@@ -298,7 +298,7 @@ describe('e2e flow snapshot — warehouse-source', () => {
       screen: 'warehouse-intro',
       action: 'confirm_setup',
     });
-    expect(trace.at(-1)?.action).toBe('keep_skills');
+    expect(trace.at(-1)?.action).toBe('dismiss_outro');
     expect(trace.length).toBeLessThan(40);
   });
 });

e2e-harness/__tests__/e2e-profile-ask.test.ts

@@ -543,9 +543,8 @@ describe('warehouse-source profile', () => {
     expect(warehouse.askAnswers?.length).toBeGreaterThan(0);
   });
 
-  it('keeps the task notice and deletes installed skills', () => {
+  it('keeps the task notice', () => {
     expect(warehouse.notice).toBe('keep');
-    expect(warehouse.skills).toBe('delete');
     expect(warehouse.mcp).toBe('skip');
   });
 

e2e-harness/__tests__/e2e-result.test.ts

@@ -410,7 +410,7 @@ describe('buildE2eResult', () => {
     newDeps: ['posthog-node'],
     envFile: '/app/.env',
     screenPath: ['warehouse-intro', 'run'],
-    skillsComplete: true,
+    flowComplete: true,
   };
 
   function build(recorder = new E2eRunRecorder()) {
@@ -454,7 +454,7 @@ describe('buildE2eResult', () => {
         'reportFile',
         'runPhase',
         'screenPath',
-        'skillsComplete',
+        'flowComplete',
         'taskOutcomes',
         'tasks',
         'unansweredAsks',
@@ -466,24 +466,24 @@ describe('buildE2eResult', () => {
     expect(build()).toMatchObject(base);
   });
 
-  it('replaces an outro result only when the final skills decision is written', () => {
+  it('replaces an outro result only when the end of the flow is written', () => {
     const directory = fs.mkdtempSync(path.join(os.tmpdir(), 'wizard-result-'));
     const file = path.join(directory, 'result.json');
-    let skillsComplete = false;
+    let flowComplete = false;
     const write = createE2eResultWriter(file, () => ({
       ...build(),
-      skillsComplete,
+      flowComplete,
     }));
     const written = () => JSON.parse(fs.readFileSync(file, 'utf8'));
 
     try {
       write();
-      expect(written()).toMatchObject({ skillsComplete: false });
-      skillsComplete = true;
+      expect(written()).toMatchObject({ flowComplete: false });
+      flowComplete = true;
       write();
-      expect(written()).toMatchObject({ skillsComplete: false });
+      expect(written()).toMatchObject({ flowComplete: false });
       write(true);
-      expect(written()).toMatchObject({ skillsComplete: true });
+      expect(written()).toMatchObject({ flowComplete: true });
     } finally {
       fs.rmSync(directory, { recursive: true, force: true });
     }
@@ -591,7 +591,7 @@ describe('security rail — no answer value reaches the payload', () => {
           newDeps: [],
           envFile: null,
           screenPath: [],
-          skillsComplete: true,
+          flowComplete: true,
         },
         recorder,
         session: { frameworkContext: {}, outroData: null },