wiggum-cli 0.13.2 → 0.15.0

package/README.md

@@ -9,6 +9,7 @@
 <p align="center">
   <a href="https://www.npmjs.com/package/wiggum-cli"><img src="https://img.shields.io/npm/v/wiggum-cli?color=F8DB27&labelColor=1a1a1a&style=flat-square" alt="npm"></a>
   <a href="https://www.npmjs.com/package/wiggum-cli"><img src="https://img.shields.io/npm/dm/wiggum-cli?color=F8DB27&labelColor=1a1a1a&style=flat-square" alt="downloads"></a>
+  <a href="https://github.com/federiconeri/wiggum-cli/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/federiconeri/wiggum-cli/ci.yml?branch=main&color=F8DB27&labelColor=1a1a1a&style=flat-square&label=CI" alt="CI"></a>
   <a href="https://github.com/federiconeri/wiggum-cli/stargazers"><img src="https://img.shields.io/github/stars/federiconeri/wiggum-cli?color=F8DB27&labelColor=1a1a1a&style=flat-square" alt="stars"></a>
   <a href="https://github.com/federiconeri/wiggum-cli/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-MIT%20+%20Commons%20Clause-F8DB27?labelColor=1a1a1a&style=flat-square" alt="license"></a>
   <img src="https://img.shields.io/node/v/wiggum-cli?color=F8DB27&labelColor=1a1a1a&style=flat-square" alt="node">
@@ -18,16 +19,22 @@
   <a href="#-quick-start">Quick Start</a> ·
   <a href="#-how-it-works">How It Works</a> ·
   <a href="https://wiggum.app">Website</a> ·
+  <a href="https://wiggum.app/blog">Blog</a> ·
+  <a href="https://wiggum.app/pricing">Pricing</a> ·
   <a href="https://github.com/federiconeri/wiggum-cli/issues">Issues</a>
 </p>
 
+<p align="center">
+  <img src=".github/screenshot.png" alt="Wiggum TUI — spec generation and autonomous coding loop" width="800">
+</p>
+
 ---
 
 ## What is Wiggum?
 
 Wiggum is an **AI agent** that plugs into any codebase and makes it ready for autonomous feature development — no configuration, no boilerplate.
 
-It works in two phases. First, **Wiggum itself is the agent**: it scans your project, detects your stack, and runs an AI-guided interview to produce detailed specs, prompts, and scripts — all tailored to your codebase. Then it delegates the actual coding to [Claude Code](https://docs.anthropic.com/en/docs/claude-code), Codex, or any CLI-based coding agent, running an autonomous **implement → test → fix** loop until the feature ships.
+It works in two phases. First, **Wiggum itself is the agent**: it scans your project, detects your stack, and runs an AI-guided interview to produce detailed specs, prompts, and scripts — all tailored to your codebase. Then it delegates the actual coding to [Claude Code](https://docs.anthropic.com/en/docs/claude-code) or any CLI-based coding agent, running an autonomous **implement → test → fix** loop until the feature ships.
 
 Plug & play. Point it at a repo. It figures out the rest.
 
@@ -41,7 +48,7 @@ Plug & play. Point it at a repo. It figures out the rest.
   │  plug&play   prompts     guides    until done        │
   │                            │    │                    │
   └────────────────────────────┘    └────────────────────┘
-       runs in your terminal          Claude Code / Codex
+       runs in your terminal          Claude Code / any agent
 ```
 
 ---
@@ -163,7 +170,8 @@ $ wiggum
 │   ├── PROMPT_e2e.md         # E2E testing
 │   ├── PROMPT_verify.md      # Verification
 │   ├── PROMPT_review_manual.md  # PR review (manual - stop at PR)
-│   └── PROMPT_review_auto.md    # PR review (auto - review + merge)
+│   ├── PROMPT_review_auto.md    # PR review (auto - review, no merge)
+│   └── PROMPT_review_merge.md   # PR review (merge - review + auto-merge)
 ├── guides/
 │   ├── AGENTS.md             # Agent instructions (CLAUDE.md)
 │   ├── FRONTEND.md           # Frontend patterns
@@ -218,8 +226,9 @@ Run the autonomous development loop.
 | `--worktree` | Git worktree isolation (parallel features) |
 | `--resume` | Resume an interrupted loop |
 | `--model <model>` | Claude model (`opus`, `sonnet`) |
-| `--max-iterations <n>` | Max iterations (default: 50) |
-| `--max-e2e-attempts <n>` | Max E2E retries (default: 3) |
+| `--max-iterations <n>` | Max iterations (default: 10) |
+| `--max-e2e-attempts <n>` | Max E2E retries (default: 5) |
+| `--review-mode <mode>` | `manual` (stop at PR), `auto` (review, no merge), or `merge` (review + merge). Default: `manual` |
 
 </details>
 
@@ -307,6 +316,16 @@ npm test
 
 ---
 
+## 📖 Learn More
+
+- [What Is Wiggum CLI?](https://wiggum.app/blog/what-is-wiggum-cli) — Overview of the autonomous coding agent
+- [What Is the Ralph Loop?](https://wiggum.app/blog/what-is-the-ralph-loop) — Deep dive into the Ralph loop methodology
+- [Wiggum vs Bash Scripts](https://wiggum.app/blog/wiggum-vs-ralph-wiggum-scripts) — Why spec generation matters
+- [Roadmap](https://wiggum.app/roadmap) — What's coming next
+- [Changelog](https://wiggum.app/changelog) — Release history
+
+---
+
 ## 📄 License
 
 **MIT + Commons Clause** — see [LICENSE](LICENSE).

package/dist/ai/providers.js

@@ -34,21 +34,25 @@ export const KNOWN_API_KEYS = [
  */
 export const AVAILABLE_MODELS = {
     anthropic: [
-        { value: 'claude-opus-4-5-20250514', label: 'Claude Opus 4.5', hint: 'most capable' },
-        { value: 'claude-sonnet-4-5-20250514', label: 'Claude Sonnet 4.5', hint: 'recommended' },
-        { value: 'claude-haiku-4-5-20250514', label: 'Claude Haiku 4.5', hint: 'fastest' },
+        { value: 'claude-opus-4-6', label: 'Claude Opus 4.6', hint: 'most capable' },
+        { value: 'claude-sonnet-4-5-20250929', label: 'Claude Sonnet 4.5', hint: 'recommended' },
+        { value: 'claude-haiku-4-5-20251001', label: 'Claude Haiku 4.5', hint: 'fastest' },
     ],
     openai: [
-        { value: 'gpt-5.1', label: 'GPT-5.1', hint: 'most capable' },
-        { value: 'gpt-5.1-codex-max', label: 'GPT-5.1 Codex Max', hint: 'best for code' },
+        { value: 'gpt-5.2', label: 'GPT-5.2', hint: 'most capable' },
+        { value: 'gpt-5.2-codex', label: 'GPT-5.2 Codex', hint: 'best for code' },
+        { value: 'gpt-5.1', label: 'GPT-5.1', hint: 'previous gen' },
+        { value: 'gpt-5.1-codex-max', label: 'GPT-5.1 Codex Max', hint: 'previous codex' },
         { value: 'gpt-5-mini', label: 'GPT-5 Mini', hint: 'fastest' },
     ],
     openrouter: [
-        { value: 'google/gemini-3-pro-preview', label: 'Gemini 3 Pro', hint: 'Google' },
-        { value: 'google/gemini-3-flash-preview', label: 'Gemini 3 Flash', hint: 'fast' },
-        { value: 'deepseek/deepseek-v3.2', label: 'DeepSeek v3.2', hint: 'efficient' },
-        { value: 'z-ai/glm-4.7', label: 'GLM 4.7', hint: 'Z-AI' },
+        { value: 'google/gemini-3-pro-preview', label: 'Gemini 3 Pro Preview', hint: 'Google' },
+        { value: 'google/gemini-3-flash-preview', label: 'Gemini 3 Flash Preview', hint: 'fast' },
+        { value: 'moonshotai/kimi-k2.5', label: 'Kimi K2.5', hint: 'Moonshot' },
+        { value: 'deepseek/deepseek-v3.2', label: 'DeepSeek V3.2', hint: 'efficient' },
         { value: 'minimax/minimax-m2.1', label: 'MiniMax M2.1', hint: 'MiniMax' },
+        { value: 'z-ai/glm-4.7', label: 'GLM 4.7', hint: 'Z-AI' },
+        { value: 'x-ai/grok-4.1-fast', label: 'Grok 4.1 Fast', hint: 'xAI' },
     ],
 };
 /**
@@ -56,17 +60,17 @@ export const AVAILABLE_MODELS = {
  * Using balanced models for good results
  */
 const DEFAULT_MODELS = {
-    anthropic: 'claude-sonnet-4-5-20250514',
-    openai: 'gpt-5.1',
+    anthropic: 'claude-sonnet-4-5-20250929',
+    openai: 'gpt-5.2',
     openrouter: 'google/gemini-3-pro-preview',
 };
 /**
  * Anthropic shorthand aliases for legacy configs
  */
 const ANTHROPIC_MODEL_ALIASES = {
-    sonnet: 'claude-sonnet-4-5-20250514',
-    opus: 'claude-opus-4-5-20250514',
-    haiku: 'claude-haiku-4-5-20250514',
+    sonnet: 'claude-sonnet-4-5-20250929',
+    opus: 'claude-opus-4-6',
+    haiku: 'claude-haiku-4-5-20251001',
 };
 /**
  * Check if a model id is an Anthropic alias (sonnet/opus/haiku)
@@ -173,6 +177,7 @@ const REASONING_MODELS = [
     'o3', 'o3-mini',
     'gpt-5', 'gpt-5.1', 'gpt-5-mini',
     'gpt-5.1-codex', 'gpt-5.1-codex-max',
+    'gpt-5.2', 'gpt-5.2-codex',
 ];
 /**
  * Check if a model is a reasoning model that doesn't support temperature

package/dist/commands/run.d.ts

@@ -8,7 +8,7 @@ export interface RunOptions {
     model?: string;
     maxIterations?: number;
     maxE2eAttempts?: number;
-    reviewMode?: 'manual' | 'auto';
+    reviewMode?: 'manual' | 'auto' | 'merge';
 }
 /**
  * Run the feature development loop for a specific feature

package/dist/commands/run.js

@@ -113,8 +113,8 @@ export async function runCommand(feature, options = {}) {
     }
     // Resolve and validate reviewMode
     const reviewMode = options.reviewMode ?? config.loop.reviewMode ?? 'manual';
-    if (reviewMode !== 'manual' && reviewMode !== 'auto') {
-        logger.error(`Invalid reviewMode '${reviewMode}'. Allowed values are 'manual' or 'auto'.`);
+    if (reviewMode !== 'manual' && reviewMode !== 'auto' && reviewMode !== 'merge') {
+        logger.error(`Invalid reviewMode '${reviewMode}'. Allowed values are 'manual', 'auto', or 'merge'.`);
         process.exit(1);
     }
     args.push('--review-mode', reviewMode);

package/dist/index.js

@@ -1,5 +1,6 @@
 import { createSessionState } from './repl/session-state.js';
 import { hasConfig, loadConfigWithDefaults } from './utils/config.js';
+import { listSpecNames } from './utils/spec-names.js';
 import { AVAILABLE_MODELS, getAvailableProvider, isAnthropicAlias } from './ai/providers.js';
 import { notifyIfUpdateAvailable } from './utils/update-check.js';
 import { renderApp } from './tui/app.js';
@@ -54,9 +55,14 @@ async function startInkTui(initialScreen = 'shell', interviewFeature) {
                 model = configuredModel;
             }
         }
-        return createSessionState(projectRoot, provider, // May be null if no API key
+        const specsDir = config
+            ? join(projectRoot, config.paths.specs)
+            : join(projectRoot, '.ralph/specs');
+        const specNames = await listSpecNames(specsDir);
+        const state = createSessionState(projectRoot, provider, // May be null if no API key
         model, undefined, // No scan result yet
         config, isInitialized);
+        return { ...state, specNames };
     }
     const initialState = await createCurrentSessionState();
     // Build interview props if starting on interview screen

package/dist/repl/session-state.d.ts

@@ -25,6 +25,8 @@ export interface SessionState {
     conversationContext?: string;
     /** Whether /init has been run in this session */
     initialized: boolean;
+    /** Cached spec names from the configured specs directory */
+    specNames?: string[];
 }
 /**
  * Create a new session state

package/dist/templates/config/ralph.config.cjs.tmpl

@@ -34,6 +34,6 @@ module.exports = {
     maxE2eAttempts: 5,
     defaultModel: 'sonnet',
     planningModel: 'opus',
-    reviewMode: 'manual', // 'manual' = stop at PR, 'auto' = review + auto-merge
+    reviewMode: 'manual', // 'manual' = stop at PR, 'auto' = review (no merge), 'merge' = review + auto-merge
   },
 };

package/dist/templates/prompts/PROMPT_review_auto.md.tmpl

@@ -95,7 +95,7 @@ fi
 ```
 
 **Handle review feedback:**
-- If Claude outputs "APPROVED" -> Proceed to Step 5 (rebase and merge)
+- If Claude outputs "APPROVED" -> Done. The PR is ready for manual merge by the user.
 - If Claude lists issues:
   1. Address each issue with code fixes
   2. Commit: `git -C {{appDir}} add -A && git -C {{appDir}} commit -m "fix($FEATURE): address review feedback"`
@@ -103,47 +103,14 @@ fi
   4. Re-run the Claude review command above
 - Max 3 review iterations before requiring manual intervention
 
-### Step 5: Rebase Before Merge (for Parallel Execution)
-Before merging, ensure branch is up-to-date with main:
-```bash
-cd {{appDir}} && git fetch origin main
-cd {{appDir}} && git rebase origin/main
-```
-
-If rebase has conflicts:
-1. Resolve conflicts in affected files
-2. `git add .` the resolved files
-3. `git rebase --continue`
-4. Re-run tests: `{{testCommand}} && {{buildCommand}}`
-
-Push rebased branch:
-```bash
-cd {{appDir}} && git push --force-with-lease origin feat/$FEATURE
-```
-
-### Step 6: Merge PR
-When Claude review is approved and branch is rebased:
-```bash
-cd {{appDir}} && gh pr merge --squash --delete-branch
-```
-
-### Step 7: Post-Merge Cleanup
-1. If using worktree, remove it:
-   ```bash
-   # Only if this feature used a worktree ({{appDir}}-$FEATURE directory exists)
-   git -C {{appDir}} worktree remove "../{{appDir}}-$FEATURE" 2>/dev/null || true
-   ```
-2. Checkout main and pull:
-   ```bash
-   git -C {{appDir}} checkout main && git -C {{appDir}} pull
-   ```
-
-Note: Spec status updates are handled in the Spec Verification phase before PR creation.
+### Step 5: Final Summary
+After review is complete (approved or max iterations reached):
+1. Post a summary comment on the PR with the review outcome
+2. Do NOT merge — the user will review and merge manually
 
 ## Rules
-- Do NOT merge without Claude Code approval
-- Address ALL review comments before merging
-- Use squash merge to keep history clean
+- Do NOT merge the PR — auto mode only reviews, the user merges
+- Address ALL review comments before marking as approved
 - If gh CLI fails, check authentication: `gh auth status`
 - Keep review conversation focused and professional
 
@@ -152,7 +119,6 @@ Note: Spec status updates are handled in the Spec Verification phase before PR c
 - **gh auth error** -> Run: `gh auth login`
 - **PR already exists** -> Use: `gh pr view` to see status
 - **Claude Code CLI not installed** -> Install: `npm install -g @anthropic-ai/claude-code`
-- **Rebase conflicts** -> Resolve carefully, re-run all tests after
 
 ## Learning Capture
 If the review revealed patterns worth remembering, append to @.ralph/LEARNINGS.md:

package/dist/templates/prompts/PROMPT_review_merge.md.tmpl

@@ -0,0 +1,163 @@
+## Context
+Study @.ralph/AGENTS.md for commands and patterns.
+Study @.ralph/specs/$FEATURE.md for feature specification.
+Study @.ralph/specs/$FEATURE-implementation-plan.md for completed tasks.
+
+## Learnings
+Read @.ralph/LEARNINGS.md for patterns from previous features.
+Capture any review feedback patterns for future iterations.
+
+## Task
+All implementation and E2E tasks are complete. Create PR, review, and merge.
+
+### Step 1: Verify Ready State
+1. Check all tasks are complete in implementation plan (no `- [ ]` items)
+2. Verify tests pass: `cd {{appDir}} && {{testCommand}}`
+3. Verify build succeeds: `cd {{appDir}} && {{buildCommand}}`
+
+If any fail, fix before proceeding.
+
+### Step 2: Check Git Status
+```bash
+cd {{appDir}} && git status
+cd {{appDir}} && git log --oneline -5
+```
+
+Ensure:
+- On branch `feat/$FEATURE`
+- All changes are committed
+- Branch is pushed to remote
+
+If uncommitted changes exist:
+```bash
+git -C {{appDir}} add -A && git -C {{appDir}} commit -m "chore($FEATURE): final cleanup"
+git -C {{appDir}} push origin feat/$FEATURE
+```
+
+### Step 3: Create PR
+Check if PR already exists:
+```bash
+cd {{appDir}} && gh pr list --head feat/$FEATURE
+```
+
+If no PR exists, create one:
+```bash
+cd {{appDir}} && gh pr create --base main --head feat/$FEATURE \
+  --title "feat($FEATURE): [read description from spec]" \
+  --body "$(cat <<'EOF'
+## Summary
+[Read from spec Purpose section]
+
+## Changes
+[Read from implementation plan - list completed phases]
+
+## Testing
+- [x] Unit/integration tests: 97 passing
+- [x] E2E tests: All scenarios passed via Playwright MCP
+- [x] Build succeeds
+
+## E2E Test Results
+[Copy from implementation plan Phase 9]
+
+Generated with Claude Code
+EOF
+)"
+```
+
+### Step 4: Request Claude Code Review
+
+Run automated code review using Claude Code CLI:
+
+```bash
+# Check if Claude Code CLI is installed
+if ! command -v claude &> /dev/null; then
+    echo "WARNING: Claude Code CLI not installed. Manual review needed."
+    cd {{appDir}} && gh pr comment --body "Manual review requested - Claude Code CLI not available. Install: https://docs.anthropic.com/en/docs/claude-code/overview"
+else
+    echo "Running Claude Code review..."
+
+    cd {{appDir}} && claude -p "You are reviewing a PR for the $FEATURE feature.
+
+Review the git diff against main and check:
+- Code quality and patterns consistency
+- Test coverage adequacy
+- Potential bugs or edge cases
+- Security concerns (injection, XSS, etc.)
+- Performance implications
+- Error handling completeness
+
+Run: git diff main
+
+Respond with:
+- APPROVED if everything looks good
+- Or list specific issues with file:line references that need to be fixed"
+fi
+```
+
+**Handle review feedback:**
+- If Claude outputs "APPROVED" -> Proceed to Step 5 (rebase and merge)
+- If Claude lists issues:
+  1. Address each issue with code fixes
+  2. Commit: `git -C {{appDir}} add -A && git -C {{appDir}} commit -m "fix($FEATURE): address review feedback"`
+  3. Push: `git -C {{appDir}} push origin feat/$FEATURE`
+  4. Re-run the Claude review command above
+- Max 3 review iterations before requiring manual intervention
+
+### Step 5: Rebase Before Merge (for Parallel Execution)
+Before merging, ensure branch is up-to-date with main:
+```bash
+cd {{appDir}} && git fetch origin main
+cd {{appDir}} && git rebase origin/main
+```
+
+If rebase has conflicts:
+1. Resolve conflicts in affected files
+2. `git add .` the resolved files
+3. `git rebase --continue`
+4. Re-run tests: `{{testCommand}} && {{buildCommand}}`
+
+Push rebased branch:
+```bash
+cd {{appDir}} && git push --force-with-lease origin feat/$FEATURE
+```
+
+### Step 6: Merge PR
+When Claude review is approved and branch is rebased:
+```bash
+cd {{appDir}} && gh pr merge --squash --delete-branch
+```
+
+### Step 7: Post-Merge Cleanup
+1. If using worktree, remove it:
+   ```bash
+   # Only if this feature used a worktree ({{appDir}}-$FEATURE directory exists)
+   git -C {{appDir}} worktree remove "../{{appDir}}-$FEATURE" 2>/dev/null || true
+   ```
+2. Checkout main and pull:
+   ```bash
+   git -C {{appDir}} checkout main && git -C {{appDir}} pull
+   ```
+
+Note: Spec status updates are handled in the Spec Verification phase before PR creation.
+
+## Rules
+- Do NOT merge without Claude Code approval
+- Address ALL review comments before merging
+- Use squash merge to keep history clean
+- If gh CLI fails, check authentication: `gh auth status`
+- Keep review conversation focused and professional
+
+## Troubleshooting
+- **gh: command not found** -> Install GitHub CLI: `brew install gh`
+- **gh auth error** -> Run: `gh auth login`
+- **PR already exists** -> Use: `gh pr view` to see status
+- **Claude Code CLI not installed** -> Install: `npm install -g @anthropic-ai/claude-code`
+- **Rebase conflicts** -> Resolve carefully, re-run all tests after
+
+## Learning Capture
+If the review revealed patterns worth remembering, append to @.ralph/LEARNINGS.md:
+- Code quality feedback -> Add under "## Anti-Patterns" or "## Patterns"
+- Common review issues -> Add under "## Anti-Patterns"
+- Good practices identified -> Add under "## Patterns"
+
+Format: `- [YYYY-MM-DD] [$FEATURE] Brief description`

package/dist/templates/scripts/feature-loop-actions.test.ts

@@ -0,0 +1,92 @@
+/**
+ * Tests for the action request JSON embedded in feature-loop.sh.tmpl
+ *
+ * The shell template contains a JSON payload written by write_action_request().
+ * These tests validate that the JSON structure matches the ActionRequest schema
+ * expected by the TUI and action-inbox helpers.
+ *
+ * Note: Full bash integration tests are out of scope. This focuses on the
+ * JSON schema validation by reading the template file directly.
+ */
+
+import { describe, it, expect } from 'vitest';
+import { readFileSync } from 'node:fs';
+import { join, dirname } from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+const TEMPLATE_PATH = join(__dirname, 'feature-loop.sh.tmpl');
+
+/**
+ * Extract the JSON block written by write_action_request() from the template.
+ * The JSON lives between `cat > "$action_file" << 'EOF'` and `EOF`.
+ */
+function extractActionRequestJson(): unknown {
+  const content = readFileSync(TEMPLATE_PATH, 'utf-8');
+
+  // Match the heredoc block: cat > ... << 'EOF' ... EOF
+  const match = content.match(/cat\s*>\s*"\$action_file"\s*<<\s*'EOF'\s*\n([\s\S]*?)\nEOF/);
+  if (!match?.[1]) {
+    throw new Error('Could not find action request JSON in feature-loop.sh.tmpl');
+  }
+
+  return JSON.parse(match[1]);
+}
+
+describe('feature-loop.sh.tmpl — action request JSON schema', () => {
+  it('parses the embedded JSON without errors', () => {
+    const parsed = extractActionRequestJson();
+    expect(parsed).toBeDefined();
+  });
+
+  it('has a non-empty string id field', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    expect(typeof parsed.id).toBe('string');
+    expect((parsed.id as string).length).toBeGreaterThan(0);
+  });
+
+  it('has a non-empty string prompt field', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    expect(typeof parsed.prompt).toBe('string');
+    expect((parsed.prompt as string).length).toBeGreaterThan(0);
+  });
+
+  it('has a choices array with at least one entry', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    expect(Array.isArray(parsed.choices)).toBe(true);
+    expect((parsed.choices as unknown[]).length).toBeGreaterThan(0);
+  });
+
+  it('each choice has a non-empty string id and label', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    const choices = parsed.choices as Array<Record<string, unknown>>;
+
+    for (const choice of choices) {
+      expect(typeof choice.id).toBe('string');
+      expect((choice.id as string).length).toBeGreaterThan(0);
+      expect(typeof choice.label).toBe('string');
+      expect((choice.label as string).length).toBeGreaterThan(0);
+    }
+  });
+
+  it('has a non-empty string default field', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    expect(typeof parsed.default).toBe('string');
+    expect((parsed.default as string).length).toBeGreaterThan(0);
+  });
+
+  it('default value matches one of the choice ids', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    const choices = parsed.choices as Array<Record<string, unknown>>;
+    const choiceIds = choices.map((c) => c.id as string);
+    expect(choiceIds).toContain(parsed.default as string);
+  });
+
+  it('all required ActionRequest fields are present', () => {
+    const parsed = extractActionRequestJson() as Record<string, unknown>;
+    expect(parsed).toHaveProperty('id');
+    expect(parsed).toHaveProperty('prompt');
+    expect(parsed).toHaveProperty('choices');
+    expect(parsed).toHaveProperty('default');
+  });
+});