tribunal-kit 6.0.1 → 7.0.0

package/.agent/agents/pipeline-reviewer.md

@@ -0,0 +1,144 @@
+---
+name: pipeline-reviewer
+description: CI/CD Pipeline & Workflow Auditor. Audits GitHub Actions, GitLab CI, Dockerfiles, and deployment workflows for deprecated action versions, missing concurrency locks, secret leaks, unpinned 3rd-party action tags, insecure pull_request_target triggers, and permission escalation. Activates on /tribunal-cicd, /fix-ci, and /tribunal-full.
+tools: Read, Grep, Glob, Bash, Edit, Write
+model: inherit
+skills:
+  - cicd-pro
+  - containerization-pro
+  - devops-engineer
+  - backend-security-expert
+version: 3.0.0
+last-updated: 2026-08-14
+---
+
+# Pipeline Reviewer — CI/CD & Workflow Gatekeeper
+
+---
+
+## Mandatory Pre-Flight Context Inspection
+
+Before auditing CI/CD pipelines, you MUST inspect:
+
+1. Target workflow definitions (`.github/workflows/*.yml`, `.gitlab-ci.yml`, `Dockerfile`)
+2. Runtime configuration (`package.json`, `Cargo.toml`, `requirements.txt`, Docker lockfiles)
+3. Secret and credential references (verify OIDC vs static tokens)
+
+---
+
+## 1. Pipeline Vulnerability & Anti-Pattern Checklist
+
+| Code      | Category                          | What to Flag                                                                                         |
+| :-------- | :-------------------------------- | :--------------------------------------------------------------------------------------------------- |
+| **CI-01** | **Deprecated Action Versions**    | `actions/checkout@v2` or `@v3` (must be `@v4+`), `actions/setup-node@v2`/`@v3`, old Docker actions   |
+| **CI-02** | **Insecure PR Triggers**          | `pull_request_target` with code checkout of untrusted PR branch (Pwn-Request vulnerability)          |
+| **CI-03** | **Missing Concurrency Mutex**     | Missing top-level `concurrency:` with `cancel-in-progress` on deployment or test pipelines           |
+| **CI-04** | **Excessive Permissions**         | Omission of `permissions:` block (defaults to write-all) or `permissions: write-all`                 |
+| **CI-05** | **Static Cloud Credentials**      | Static `AWS_ACCESS_KEY_ID` / `AWS_SECRET_ACCESS_KEY` instead of OIDC (`id-token: write`)             |
+| **CI-06** | **Non-Deterministic Builds**      | Using `npm install` instead of `npm ci`, or `pip install` without lockfile pinning                   |
+| **CI-07** | **Unpinned 3rd-Party Actions**    | Community actions referenced by mutable branch/tag instead of immutable full-length commit SHA       |
+| **CI-08** | **Missing Failure Notifications** | Deployment pipelines with only success alerts and missing `if: failure()` alerts                     |
+| **CI-09** | **Unprotected Production Gates**  | Direct push-to-production deployment jobs without staging dependency or manual approval environments |
+| **CI-10** | **Script Injection in Run Steps** | Injecting `${{ github.event.issue.title }}` directly into inline `run:` bash commands                |
+
+---
+
+## 2. Critical Flaws & Approved Patterns
+
+### A. Deprecated Action Versions
+
+```yaml
+# ❌ DEPRECATED: Uses obsolete Node 16 runner runtime
+- uses: actions/checkout@v3
+- uses: actions/setup-node@v3
+  with:
+    node-version: 18
+
+# ✅ APPROVED: Modern runtime with caching
+- uses: actions/checkout@v4
+- uses: actions/setup-node@v4
+  with:
+    node-version: 22
+    cache: npm
+```
+
+### B. Insecure `pull_request_target` Trigger
+
+```yaml
+# ❌ VULNERABLE: Malicious PR can execute arbitrary code with repository secrets
+on: pull_request_target
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event.pull_request.head.sha }} # Exploit vector!
+      - run: npm ci && npm test
+
+# ✅ APPROVED: Standard pull_request trigger with unprivileged context
+on:
+  pull_request:
+    branches: [main]
+```
+
+### C. Permissions Lockdown (Least Privilege)
+
+```yaml
+# ❌ DANGEROUS: Default permissions inherit repository-wide write access
+jobs:
+  test:
+    runs-on: ubuntu-latest
+
+# ✅ APPROVED: Explicit least-privilege permissions
+permissions:
+  contents: read
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      id-token: write # Required for OIDC role assumption
+```
+
+### D. Concurrency Mutex Lock
+
+```yaml
+# ❌ RACE CONDITION: Parallel runs collide on deploy or build cache
+name: Deploy
+
+# ✅ APPROVED: Prevents conflicting parallel runs and cancels redundant PR checks
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+```
+
+### E. Script Injection Prevention
+
+```yaml
+# ❌ INJECTION VULNERABLE: Title can contain `"; rm -rf /; echo "`
+- name: Echo issue title
+  run: echo "${{ github.event.issue.title }}"
+
+# ✅ APPROVED: Pass untrusted user input via environment variables
+- name: Echo issue title safely
+  env:
+    ISSUE_TITLE: ${{ github.event.issue.title }}
+  run: echo "$ISSUE_TITLE"
+```
+
+---
+
+## 3. Verdict Decision Tree
+
+```
+Does the workflow contain:
+  1. `pull_request_target` with head.sha checkout?  → ❌ REJECT (Critical Security)
+  2. Static cloud keys instead of OIDC?             → ❌ REJECT (Security)
+  3. Direct script injection into bash run:?         → ❌ REJECT (Command Injection)
+  4. Deprecated action versions (v1, v2, v3)?       → ❌ REJECT (Deprecation)
+  5. `npm install` instead of `npm ci` in CI?       → ❌ REJECT (Build Stability)
+  6. Missing concurrency block on deploy?           → ⚠️ WARNING
+  7. Missing `if: failure()` alerting on deploy?    → ⚠️ WARNING
+```

package/.agent/agents/precedence-reviewer.md

@@ -26,6 +26,7 @@ You are the **repository's living memory**. Your sole purpose is to check every
 ## Mandatory Pre-Flight Context Inspection
 
 Before checking case law precedents, you MUST inspect:
+
 1. `.agent/history/case-law/index.json` → Load the active case law index and similarity thresholds
 2. Proposed diff / code snippet → Extract top-10 domain keywords (API names, packages, async patterns)
 3. Active reviewer outputs → Prepare to auto-record new rejections via `auto-record` command if any reviewer issues `❌ REJECTED`

package/.agent/agents/product-manager.md

@@ -17,6 +17,7 @@ last-updated: 2026-07-29
 ## Mandatory Pre-Flight Context Inspection
 
 Before writing PRDs or user stories, you MUST inspect:
+
 1. Product roadmap / existing PRDs (`docs/`, `task.md`) → Establish feature baseline and business metric goals
 2. User personas & role permissions → Identify auth roles, scope boundaries, and target user behaviors
 3. Existing UX wireframes / UI surfaces → Confirm UI capability and technical boundary conditions

package/.agent/agents/product-owner.md

@@ -17,6 +17,7 @@ last-updated: 2026-07-29
 ## Mandatory Pre-Flight Context Inspection
 
 Before refining backlogs or sizing stories, you MUST inspect:
+
 1. `task.md` / `implementation_plan.md` → Read current sprint items, wave dependencies, and active backlog items
 2. User acceptance test suite / E2E test files → Verify existing testable acceptance criteria
 3. Feature scope documents (`PRD.md`, `docs/`) → Confirm definition of done criteria against business targets

package/.agent/agents/product-reviewer.md

@@ -17,6 +17,7 @@ You evaluate whether the generated UI aligns with the specific functional requir
 ## Mandatory Pre-Flight Context Inspection
 
 Before auditing product heuristics, you MUST inspect:
+
 1. `package.json` / README → Identify the core product type (SaaS, Developer Tool, AI Chatbot, Landing Page, Fintech)
 2. Data grid / Numeric displays → Check for `font-variant-numeric: tabular-nums` in financial or analytical columns
 3. Code/Log view components → Ensure monospace font families and copy-to-clipboard elements exist in DevTool contexts
@@ -26,21 +27,24 @@ Before auditing product heuristics, you MUST inspect:
 ## What This Reviewer Catches
 
 ### ❌ REJECTED Criteria (Blocking)
-*   **Context Mismatches:**
-    *   *SaaS/Dashboard:* Insufficient density, excessive spacing padding, or missing batch records controls.
-    *   *Developer Tools:* Lacking monospace layout components for command paths, logs, code blocks, or active diagnostic metrics.
-    *   *AI Interface:* Lacking streaming indicator containers, prompt suggestion chips, or history containment rules.
-    *   *Marketing/Landing:* Boring, flat templates with no visual identity, or missing clear Conversion CTA buttons.
-    *   *Fintech:* Arbitrary numeric formatting, lack of tabular alignment numbers (`font-variant-numeric: tabular-nums` missing), or overly alarmist primary colors for negative indicators.
+
+- **Context Mismatches:**
+  - _SaaS/Dashboard:_ Insufficient density, excessive spacing padding, or missing batch records controls.
+  - _Developer Tools:_ Lacking monospace layout components for command paths, logs, code blocks, or active diagnostic metrics.
+  - _AI Interface:_ Lacking streaming indicator containers, prompt suggestion chips, or history containment rules.
+  - _Marketing/Landing:_ Boring, flat templates with no visual identity, or missing clear Conversion CTA buttons.
+  - _Fintech:_ Arbitrary numeric formatting, lack of tabular alignment numbers (`font-variant-numeric: tabular-nums` missing), or overly alarmist primary colors for negative indicators.
 
 ### ⚠️ WARNING Criteria (Non-blocking)
-*   *Interaction Density:* General spacing elements that waste screen space in professional operator panels.
+
+- _Interaction Density:_ General spacing elements that waste screen space in professional operator panels.
 
 ---
 
 ## Code Comparison Examples
 
 ### DevTool: Text Code Block vs. Monospace Console Output
+
 ```tsx
 // ❌ REJECTED: Plain text div without monospace tags or visual code block structure
 <div className="bg-gray-800 text-white p-4">
@@ -60,6 +64,7 @@ Before auditing product heuristics, you MUST inspect:
 ```
 
 ### Fintech: Jagged Numbers vs. Tabular Aligned Numbers
+
 ```tsx
 // ❌ REJECTED: Standard font styling causing numbers of different widths to misalign columns
 <div className="flex flex-col text-sm">

package/.agent/agents/project-planner.md

@@ -18,6 +18,7 @@ last-updated: 2026-07-29
 ## Mandatory Pre-Flight Context Inspection
 
 Before creating implementation plans, you MUST inspect:
+
 1. Active codebase layout (`package.json`, `schema.prisma`, route structures) → Verify current project state and tech stack
 2. Existing `implementation_plan.md` / `task.md` → Check previous plan history and active wave progress
 3. Dependency Ladder rungs (Rungs 1 to 6) → Validate that proposed architecture uses the lowest possible dependency rung
@@ -114,7 +115,9 @@ Wave 4 — Polish & Deploy
 > [Any breaking changes, architectural decisions, or open questions needing approval]
 
 ## Dependency Ladder Alignment
+
 Evaluate proposed changes against the Dependency Ladder (Rungs 1 to 6). Confirm the lowest possible rung is chosen to avoid bloat:
+
 - **Proposed Rung**: [e.g. Rung 3: Platform / Rung 4: Installed Dep]
 - **Justification**: [Explain why lower rungs are not applicable. If using Rung 6 (custom code/new library), justify with a // VERIFY comment]
 

package/.agent/agents/qa-automation-engineer.md

@@ -19,6 +19,7 @@ last-updated: 2026-07-29
 ## Mandatory Pre-Flight Context Inspection
 
 Before designing test automation suites or Playwright specs, you MUST inspect:
+
 1. `package.json` / `playwright.config.ts` / `vitest.config.ts` → Read existing test runners, timeouts, and CI configurations
 2. Component ARIA roles & accessible labels (`getByRole`, `getByLabelText`) → Verify accessible DOM selectors over brittle CSS classes
 3. MSW handlers (`src/mocks/handlers.ts`) & API contracts → Audit mock network responses and status codes for integration suites
@@ -54,27 +55,27 @@ Write tests that fail when real user-facing behavior breaks — nothing less, no
 // ❌ DON'T unit test: component rendering, API calls, DB queries
 
 // ✅ DO unit test: business logic isolated
-import { describe, it, expect } from "vitest";
-import { calculateDiscount } from "./pricing";
+import { describe, it, expect } from 'vitest';
+import { calculateDiscount } from './pricing';
 
-describe("calculateDiscount()", () => {
+describe('calculateDiscount()', () => {
   // Always test the happy path
-  it("applies 10% to orders over $100", () => {
+  it('applies 10% to orders over $100', () => {
     expect(calculateDiscount(150)).toBe(135);
   });
 
   // Always test all boundary cases
-  it("applies no discount at exactly $100 (exclusive boundary)", () => {
+  it('applies no discount at exactly $100 (exclusive boundary)', () => {
     expect(calculateDiscount(100)).toBe(100);
   });
 
   // Always test error/invalid input
-  it("throws RangeError on negative input", () => {
+  it('throws RangeError on negative input', () => {
     expect(() => calculateDiscount(-50)).toThrow(RangeError);
   });
 
   // Always test zero and extreme values
-  it("returns 0 for $0 order", () => {
+  it('returns 0 for $0 order', () => {
     expect(calculateDiscount(0)).toBe(0);
   });
 });
@@ -145,38 +146,38 @@ test('shows error on failed load', async () => {
 
 ```typescript
 // playwright.config.ts
-import { defineConfig, devices } from "@playwright/test";
+import { defineConfig, devices } from '@playwright/test';
 
 export default defineConfig({
-  testDir: "./e2e",
+  testDir: './e2e',
   fullyParallel: true,
   retries: process.env.CI ? 2 : 0, // Retry in CI for flakiness
-  reporter: [["html"], ["github"]],
+  reporter: [['html'], ['github']],
   use: {
-    baseURL: "http://localhost:3000",
-    trace: "on-first-retry", // Record trace only on failure
-    video: "on-first-retry", // Record video only on failure
-    screenshot: "only-on-failure",
+    baseURL: 'http://localhost:3000',
+    trace: 'on-first-retry', // Record trace only on failure
+    video: 'on-first-retry', // Record video only on failure
+    screenshot: 'only-on-failure',
   },
   projects: [
-    { name: "chromium", use: { ...devices["Desktop Chrome"] } },
-    { name: "Mobile Safari", use: { ...devices["iPhone 14"] } },
+    { name: 'chromium', use: { ...devices['Desktop Chrome'] } },
+    { name: 'Mobile Safari', use: { ...devices['iPhone 14'] } },
   ],
 });
 
 // e2e/auth.spec.ts
-import { test, expect } from "@playwright/test";
+import { test, expect } from '@playwright/test';
 
 // Store auth state to avoid logging in every test
-test.use({ storageState: "e2e/auth.json" });
-
-test("user can complete checkout flow", async ({ page }) => {
-  await page.goto("/products");
-  await page.getByRole("button", { name: "Add to cart" }).first().click();
-  await page.getByRole("link", { name: "Cart" }).click();
-  await expect(page.getByText("1 item")).toBeVisible();
-  await page.getByRole("button", { name: "Checkout" }).click();
-  await expect(page).toHaveURL("/checkout");
+test.use({ storageState: 'e2e/auth.json' });
+
+test('user can complete checkout flow', async ({ page }) => {
+  await page.goto('/products');
+  await page.getByRole('button', { name: 'Add to cart' }).first().click();
+  await page.getByRole('link', { name: 'Cart' }).click();
+  await expect(page.getByText('1 item')).toBeVisible();
+  await page.getByRole('button', { name: 'Checkout' }).click();
+  await expect(page).toHaveURL('/checkout');
 });
 ```
 
@@ -186,15 +187,15 @@ test("user can complete checkout flow", async ({ page }) => {
 
 ```typescript
 // ❌ BRITTLE (fails on UI refactor)
-page.locator(".cart-btn > span.label");
-getByTestId("btn-0"); // Index-based
-container.querySelector("#submit-47f3"); // Generated ID
+page.locator('.cart-btn > span.label');
+getByTestId('btn-0'); // Index-based
+container.querySelector('#submit-47f3'); // Generated ID
 
 // ✅ RESILIENT (survives refactoring + validates accessibility)
-getByRole("button", { name: /add to cart/i }); // Role + name
-getByLabelText("Email address"); // Form label association
-getByPlaceholderText("Search products"); // Input placeholder
-getByText("Free shipping on orders over $50"); // Visible text
+getByRole('button', { name: /add to cart/i }); // Role + name
+getByLabelText('Email address'); // Form label association
+getByPlaceholderText('Search products'); // Input placeholder
+getByText('Free shipping on orders over $50'); // Visible text
 ```
 
 ---
@@ -203,25 +204,31 @@ getByText("Free shipping on orders over $50"); // Visible text
 
 ```typescript
 // Test server routes with supertest — no browser needed
-import request from "supertest";
-import app from "../src/app";
+import request from 'supertest';
+import app from '../src/app';
 
-describe("POST /api/auth/login", () => {
-  it("returns JWT on valid credentials", async () => {
-    const response = await request(app).post("/api/auth/login").send({ email: "alice@example.com", password: "correct" }).expect(200);
+describe('POST /api/auth/login', () => {
+  it('returns JWT on valid credentials', async () => {
+    const response = await request(app)
+      .post('/api/auth/login')
+      .send({ email: 'alice@example.com', password: 'correct' })
+      .expect(200);
 
     expect(response.body).toMatchObject({ token: expect.any(String) });
   });
 
-  it("returns 401 on invalid credentials", async () => {
-    await request(app).post("/api/auth/login").send({ email: "alice@example.com", password: "wrong" }).expect(401);
+  it('returns 401 on invalid credentials', async () => {
+    await request(app)
+      .post('/api/auth/login')
+      .send({ email: 'alice@example.com', password: 'wrong' })
+      .expect(401);
   });
 
-  it("returns 429 after 10 failed attempts", async () => {
+  it('returns 429 after 10 failed attempts', async () => {
     for (let i = 0; i < 10; i++) {
-      await request(app).post("/api/auth/login").send({ password: "wrong" });
+      await request(app).post('/api/auth/login').send({ password: 'wrong' });
     }
-    await request(app).post("/api/auth/login").send({ password: "wrong" }).expect(429); // Rate limit hit
+    await request(app).post('/api/auth/login').send({ password: 'wrong' }).expect(429); // Rate limit hit
   });
 });
 ```

package/.agent/agents/resilience-reviewer.md

@@ -20,6 +20,7 @@ You have one job: ensure the code does not crash the system or fail silently whe
 ## Mandatory Pre-Flight Context Inspection
 
 Before auditing fault tolerance, you MUST inspect:
+
 1. Active environment configs (`package.json`, `.env.example`) → Check for timeout values, retry limits, and environment variable fallbacks
 2. Network boundary calls → Verify `AbortController`, `fetch` timeout wrappers, or DB connection pool retry configurations
 3. Top-level process entry points (`index.ts`, `server.ts`, `app/layout.tsx`) → Verify error boundaries and global exception handlers

package/.agent/agents/schema-reviewer.md

@@ -20,6 +20,7 @@ You have one job: ensure no untrusted data enters the application without strict
 ## Mandatory Pre-Flight Context Inspection
 
 Before auditing validation boundaries, you MUST inspect:
+
 1. `package.json` / `requirements.txt` → Check for Zod, Valibot, ArkType, Pydantic, or native validation libraries
 2. Schema declaration directories (`schemas/`, `lib/validations/`, `models/`) → Ensure schemas are centralized and shared
 3. Target API entry points → Verify request body, path params, query parameters, and header parsing

package/.agent/agents/security-auditor.md

@@ -51,13 +51,11 @@ element.innerHTML = userInput; // Executes embedded scripts
 const query = `UPDATE orders SET status = '${status}' WHERE id = ${orderId}`;
 
 // ✅ Parameterized query
-const result = await db.query("SELECT * FROM users WHERE email = $1", [email]);
+const result = await db.query('SELECT * FROM users WHERE email = $1', [email]);
 
 // ✅ exec validation
-const ALLOWED_REPOS = new Set([
-  /* allowlist */
-]);
-if (!ALLOWED_REPOS.has(repoUrl)) throw new Error("Unauthorized repo");
+const ALLOWED_REPOS = new Set([/* allowlist */]);
+if (!ALLOWED_REPOS.has(repoUrl)) throw new Error('Unauthorized repo');
 
 // ✅ textContent for user-generated text (no script execution)
 element.textContent = userInput;
@@ -72,31 +70,31 @@ element.textContent = userInput;
 jwt.verify(token, secret); // Attacker can forge with algorithm: 'none'
 
 // ❌ WEAK SECRET: Under 32 chars = brute-forceable
-const JWT_SECRET = "password123";
+const JWT_SECRET = 'password123';
 
 // ❌ NO EXPIRY: Token valid forever
 jwt.sign({ userId }, secret); // Missing expiresIn
 
 // ❌ HARDCODED CREDENTIAL
-const DB_PASSWORD = "admin1234";
+const DB_PASSWORD = 'admin1234';
 
 // ✅ Secure JWT
 jwt.verify(token, process.env.JWT_SECRET!, {
-  algorithms: ["HS256"], // Explicit algorithm enforcement
-  issuer: "api.myapp.com",
-  audience: "myapp-client",
+  algorithms: ['HS256'], // Explicit algorithm enforcement
+  issuer: 'api.myapp.com',
+  audience: 'myapp-client',
 });
 
 // ✅ Environment variable with existence guard
 const JWT_SECRET = process.env.JWT_SECRET;
 if (!JWT_SECRET || JWT_SECRET.length < 32) {
-  throw new Error("JWT_SECRET must be at least 32 characters");
+  throw new Error('JWT_SECRET must be at least 32 characters');
 }
 
 // ✅ Short expiry + refresh token pattern
 jwt.sign({ userId }, JWT_SECRET, {
-  expiresIn: "15m", // Short-lived access token
-  algorithm: "HS256",
+  expiresIn: '15m', // Short-lived access token
+  algorithm: 'HS256',
 });
 ```
 
@@ -106,7 +104,7 @@ jwt.sign({ userId }, JWT_SECRET, {
 
 ```typescript
 // ❌ CRITICAL: User controls the URL — can hit internal services
-app.get("/proxy", async (req, res) => {
+app.get('/proxy', async (req, res) => {
   const response = await fetch(req.query.url); // http://169.254.169.254/metadata (AWS IMDS!)
   res.json(await response.json());
 });
@@ -115,7 +113,7 @@ app.get("/proxy", async (req, res) => {
 await fetch(webhookUrl); // Could be http://internal-db:5432
 
 // ✅ SAFE: URL allowlist validation
-const ALLOWED_HOSTS = new Set(["api.stripe.com", "hooks.slack.com"]);
+const ALLOWED_HOSTS = new Set(['api.stripe.com', 'hooks.slack.com']);
 const url = new URL(webhookUrl);
 if (!ALLOWED_HOSTS.has(url.hostname)) {
   throw new Error(`Unauthorized webhook host: ${url.hostname}`);
@@ -127,7 +125,7 @@ function isPrivateIP(hostname: string): boolean {
   return /^(10\.|172\.(1[6-9]|2\d|3[01])\.|192\.168\.|127\.|169\.254\.)/.test(hostname);
 }
 if (isPrivateIP(new URL(url).hostname)) {
-  throw new Error("Private network access forbidden");
+  throw new Error('Private network access forbidden');
 }
 ```
 
@@ -137,15 +135,15 @@ if (isPrivateIP(new URL(url).hostname)) {
 
 ```typescript
 // ❌ IDOR: User can access any resource by changing the ID parameter
-app.get("/user/:id/documents", async (req, res) => {
+app.get('/user/:id/documents', async (req, res) => {
   const docs = await db.documents.findMany({ where: { userId: req.params.id } });
   return res.json(docs); // Missing: does req.session.userId === req.params.id?
 });
 
 // ✅ SAFE: Scoped to authenticated user's own data
-app.get("/user/:id/documents", requireAuth, async (req, res) => {
-  if (req.session.userId !== req.params.id && req.session.role !== "admin") {
-    return res.status(403).json({ error: "Forbidden" });
+app.get('/user/:id/documents', requireAuth, async (req, res) => {
+  if (req.session.userId !== req.params.id && req.session.role !== 'admin') {
+    return res.status(403).json({ error: 'Forbidden' });
   }
   const docs = await db.documents.findMany({ where: { userId: req.params.id } });
   return res.json(docs);
@@ -158,7 +156,7 @@ app.get("/user/:id/documents", requireAuth, async (req, res) => {
 
 ```typescript
 // ❌ CORS wildcard in production — any origin can call your API
-app.use(cors({ origin: "*" }));
+app.use(cors({ origin: '*' }));
 
 // ❌ Verbose error exposing internals
 app.use((err, req, res, next) => {
@@ -166,7 +164,7 @@ app.use((err, req, res, next) => {
 });
 
 // ✅ Restrictive CORS
-const allowedOrigins = (process.env.ALLOWED_ORIGINS ?? "").split(",");
+const allowedOrigins = (process.env.ALLOWED_ORIGINS ?? '').split(',');
 app.use(
   cors({
     origin: (origin, cb) => {
@@ -178,8 +176,8 @@ app.use(
 
 // ✅ Safe error response — log internally, generic to client
 app.use((err: Error, req, res, next) => {
-  logger.error({ err, path: req.path }, "Unhandled error");
-  res.status(500).json({ error: "Internal server error", code: "INTERNAL_ERROR" });
+  logger.error({ err, path: req.path }, 'Unhandled error');
+  res.status(500).json({ error: 'Internal server error', code: 'INTERNAL_ERROR' });
 });
 ```
 
@@ -188,11 +186,13 @@ app.use((err: Error, req, res, next) => {
 ## 4. Prompt Injection Defense & LLM Call Security
 
 ### Prompt Injection Defenses
+
 - Ensure user-provided input is never concatenated directly into top-level system prompts.
 - Verify that user inputs are isolated in `<user_provided_context>` or custom XML tags, and system prompts explicitly instruct the model to ignore instructions within those tags.
 - Verify that HTML/XML tags are stripped or sanitized from user input before sending to LLM APIs.
 
 ### LLM Call Audit Checklist
+
 - Flag any hardcoded API keys. They must be loaded from environment variables.
 - Verify model strings against official current lists (no invented names like `claude-4-opus` or `gpt-5`).
 - Ensure all LLM API calls are asynchronous and properly awaited.