Overhaul Testsmith prompt (#652)

* Overhaul Testsmith prompt

Teach Testsmith to design permanent cases from success_criteria with
risk prioritization, a setup/action/expect template, and Corbits report
shape. Keep blinders on design-only — refuse Tester/Builder drift despite
mounted writes; no gaas twin.

Closes CL-7033

* Format Testsmith director package with Prettier

Author: TheGreatAxios

src/agent/directors/testsmith/package.test.ts

@@ -11,10 +11,67 @@ describe("testsmithPackage", () => {
     expect(testsmithPackage.systemPrompt.startsWith("Placeholder")).toBe(false);
   });
 
-  test("systemPrompt states PRIMARY INTENT for test design", () => {
-    expect(testsmithPackage.systemPrompt).toContain("PRIMARY INTENT");
-    expect(testsmithPackage.systemPrompt).toMatch(/test strategy|test cases|design/i);
-    expect(testsmithPackage.systemPrompt).toMatch(/do not implement|not implement/i);
+  test("systemPrompt identity is Testsmith / TestsmithDirector", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).toMatch(/TestsmithDirector \(Testsmith\)/);
+    expect(p).toContain("PRIMARY INTENT");
+    expect(p).toMatch(/permanent test/i);
+  });
+
+  test("systemPrompt teaches design-in-report workflow and case template", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).toMatch(/Design-in-report workflow/i);
+    expect(p).toMatch(/Blinders on|BLINDERS ON/i);
+    expect(p).toContain("success_criteria");
+    expect(p).toMatch(/Case template/i);
+    expect(p).toMatch(/\*\*Setup\*\*/);
+    expect(p).toMatch(/\*\*Action\*\*/);
+    expect(p).toMatch(/\*\*Expect\*\*/);
+    expect(p).toMatch(/what not to test/i);
+    expect(p).toMatch(/unit \| integration \| e2e/);
+  });
+
+  test("systemPrompt teaches risk prioritization", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).toMatch(/Risk prioritization/i);
+    expect(p).toMatch(/Cover first/i);
+    expect(p).toMatch(/Defer or omit/i);
+  });
+
+  test("systemPrompt is design lane only (not Tester / Builder / orchestrator)", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).toMatch(/Do not become Builder/i);
+    expect(p).toMatch(/that is Tester/i);
+    expect(p).toMatch(/do not use them/i);
+    expect(p).toMatch(/fleet orchestration/i);
+    expect(p).toMatch(/DONE GATE/i);
+    expect(p).toMatch(/Hand off/i);
+  });
+
+  test("systemPrompt states Corbits report shape", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).toContain("## Summary");
+    expect(p).toContain("## Findings");
+    expect(p).toContain("## Blockers");
+    expect(p).toContain("## Paths");
+    expect(p).toMatch(/Corbits report shape/i);
+  });
+
+  test("systemPrompt has no tool-schema restatement or fake caps", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).not.toMatch(/parameters?:/i);
+    expect(p).not.toMatch(/fan-out/i);
+    expect(p).not.toMatch(/at most \d+/i);
+    expect(p).not.toMatch(/turn budget/i);
+    expect(p).not.toMatch(/scheduler/i);
+  });
+
+  test("systemPrompt is not a gaasbot twin", () => {
+    const p = testsmithPackage.systemPrompt;
+    expect(p).not.toMatch(/Gaasbot/i);
+    expect(p).not.toMatch(/risk counsel/i);
+    expect(p).not.toMatch(/ship-with-note/i);
+    expect(p).not.toMatch(/filed-for-later/i);
   });
 
   test("spawn.maySpawn is false (leaf)", () => {
@@ -33,14 +90,15 @@ describe("testsmithPackage", () => {
     expect(testsmithPackage.modelRole).toBe("test");
   });
 
-  test("primaryIntent is design-only and not primary verifier", () => {
-    expect(testsmithPackage.primaryIntent).toMatch(/design/i);
+  test("primaryIntent is permanent-design and not primary verifier", () => {
+    expect(testsmithPackage.primaryIntent).toMatch(/permanent test cases/i);
     expect(testsmithPackage.primaryIntent).toMatch(/not.*verifier|do not run as primary verifier/i);
   });
 
-  test("outOfLane refuses product implement and runtime verify role", () => {
+  test("outOfLane refuses product implement, verifier role, and landing tests", () => {
     const joined = testsmithPackage.outOfLane.join(" ");
     expect(joined).toMatch(/implement/i);
     expect(joined).toMatch(/verifier|tester/i);
+    expect(joined).toMatch(/landing test/i);
   });
 });

src/agent/directors/testsmith/package.ts

@@ -2,34 +2,82 @@ import type { DirectorPackage } from "../types.js";
 import { REVIEW_TOOLS } from "../tool-sets.js";
 
 /**
- * Testsmith: test design specialist — strategy and cases only; never implements product
- * and is not the runtime verifier (that is tester).
+ * Testsmith leaf (CL-7033).
+ * Design permanent test strategy and cases in the report — never implement product,
+ * never replace Tester as the runtime verifier.
  */
 export const testsmithPackage: DirectorPackage = {
   id: "testsmith",
   primaryIntent:
-    "Design test strategy and cases; do not implement product; do not run as primary verifier",
+    "Design permanent test cases; do not implement product; do not run as primary verifier",
   outOfLane: [
     "implementing product code",
     "shipping features",
     "acting as primary runtime verifier (tester)",
     "fixing failing product code",
+    "landing test files as the implementer",
     "orchestration",
   ],
-  description: "Test design specialist — strategy and cases in the report only",
-  systemPrompt: `You are TestsmithDirector, a specialist in Corbits Code.
+  description: "Test design specialist — permanent cases in the report only",
+  systemPrompt: `You are TestsmithDirector (Testsmith), a specialist in Corbits Code.
 
-PRIMARY INTENT: design test strategy and test cases for the brief. Produce clear, agent-ready coverage plans. Do not implement product code. Do not act as the primary runtime verifier (that is Tester).
+PRIMARY INTENT: design permanent test strategy and cases for the brief. Produce agent-ready coverage the suite should keep. Do not implement product code. Do not act as the primary runtime verifier (that is Tester). Do not become Builder.
 
-Design in the report. Prefer:
-- risk-based coverage and acceptance criteria from the brief
-- unit / integration / e2e boundaries when relevant
-- concrete cases: setup, action, expected result, edge/failure modes
-- what not to test and why
+You are the test-design lane only — not Tester, not Builder, not Counsel, not an orchestrator. Do not spawn specialists. Write tools are mounted with no path lock — do not use them. Leave product and test-file edits to Builder; leave suite/repro execution to Tester.
 
-OUT OF LANE: fixing production code, becoming the implementer, running the full verify-and-fix loop, fleet orchestration.
+BLINDERS ON: Design from the brief's success_criteria / acceptance criteria and stated risks — not from "whatever the code does today." Read/search only to ground paths, public APIs, and existing suite shape. Do not soften cases to match current buggy behavior. Stay on this brief; do not wander into peer work or fleet orchestration.
 
-Read and search the codebase to ground the design; do not mutate product code.`,
+# Design-in-report workflow
+
+1. Map every success_criteria item to concrete permanent cases (or Blockers if you cannot).
+2. Rank by risk: correctness/data integrity and user-visible breaks first; then API contract and regression of known failure modes; defer style theater and impossible paths.
+3. Name the boundary for each case: unit | integration | e2e — pick the cheapest layer that can prove the claim.
+4. Write each case with the template below. Prefer a few sharp permanent cases over a fog of speculative ones.
+5. Explicitly list what not to test and why (impossible paths, over-engineering theater, pure typechecker/library happy paths the project already trusts).
+6. Hand off: Builder lands the tests; Tester runs them. You design only.
+
+# Case template
+
+For every permanent case include:
+- **Name** — short, stable identifier a Builder can paste into a test title
+- **Boundary** — unit | integration | e2e
+- **Risk** — why this case earns a permanent seat (what breaks if it is missing)
+- **Setup** — fixtures, state, mocks/fakes (prefer inject clocks/I/O over sleeping/network)
+- **Action** — the single behavior under test
+- **Expect** — observable result (return, state, error shape, side effect)
+- **Edge / failure** — invalid input, missing branch, or failure mode that must stay covered
+
+# Risk prioritization
+
+Cover first:
+- Invariants that protect customers/data and stated success_criteria
+- Public API sync/async and signature contracts when the brief specifies them
+- Regression of defects the brief or Findings already named
+
+Defer or omit:
+- Speculative abstractions and defensive cases for impossible states
+- Style nits and "while we're here" coverage
+- Re-testing a well-maintained library's happy path
+
+# Corbits report shape
+
+When done, stop tooling and reply with ONLY this envelope:
+
+## Summary
+One or two sentences: strategy and coverage scope designed.
+
+## Findings
+Permanent cases (name + boundary + setup/action/expect + risk), coverage map of each success_criteria item → cases (or blocked), and what not to test with why.
+
+## Blockers
+Open questions, missing acceptance criteria, or assumptions. Write "None." if clear.
+
+## Paths
+Files/suites you read to ground the design (one per line). Write "None." if none.
+
+DONE GATE: Stop when every success_criteria item has permanent cases (or Blockers). Do not invent architecture or expand the brief after criteria are covered. If the brief is ambiguous, report Blockers — do not become Counsel or Greybeard.
+
+OUT OF LANE: implementing product or tests, becoming Tester/Builder, running the full verify-and-fix loop, fleet orchestration, architecture essays, exploration maps as primary.`,
   tools: { allow: REVIEW_TOOLS },
   spawn: { maySpawn: false },
   tier: "leaf",