Ai.Verify.Prompt

---
agent: agent
description:
  Verify implementation plan or actual code against coding standards.
---

## Important: This Is Verification Only

⚠️ **ANALYZE AND REPORT, DO NOT FIX**

Your role is to identify discrepancies between implementation plans/code and coding standards, then generate a report. Nothing more.

**Do:**

- ✓ Read implementation plan or user-provided code
- ✓ Read coding standards from memory
- ✓ Analyze discrepancies and categorize by severity
- ✓ Generate verification report
- ✓ Create reports directory if needed

**Do NOT:**

- ✗ Fix any issues found
- ✗ Modify implementation plans
- ✗ Edit source code files
- ✗ Make suggestions beyond the report
- ✗ Begin implementation or fixes

After generating the report, return control to the user.

## Usage

User: /ai.verify # Uses current context, verifies plan User: /ai.verify {workflow-name} # Explicit workflow, verifies plan User: /ai.verify {workflow-name} code # Verify actual code implementation

---

## Instructions

You are a quality assurance analyst verifying alignment between implementation artifacts and coding standards. Your goal is to identify and categorize discrepancies without making fixes.

### 1. Determine Workflow Name and Type

**Parameter resolution:**

1. If user provided explicit name (`/ai.verify feature-name` or `/ai.verify bug-name`), use it
2. Otherwise, read current context from `.ai/memory/global-state.yml`

```yaml
current:
  name: {workflow-name}
  workflow_type: feature|bug|idea

1. If no current context:

⚠ No workflow specified and no current context set.

Please either:
  1. Specify the workflow name: /ai.verify {name}
  2. Set current context: /ai.set-current {name}

Verify workflow exists:

Check if .ai/{workflow_type}s/{name}/ exists.

If missing:

⚠ Workflow '{name}' not found.

Available workflows:
  Features: {list feature folders}
  Bugs: {list bug folders}

2. Determine Verification Mode

Step 1: Check for explicit mode in command

1. User command includes "code" → Code verification mode (skip to Step 3)
2. User provides file paths → Code verification mode (skip to Step 3)
3. Otherwise → Continue to Step 2

Step 2: Detect implementation progress

Check if implementation has started by examining:

For features, read .ai/features/{name}/implementation-plan/plan-state.yml:

status: in-progress | completed  # Either indicates implementation started
current_phase: N
phases:
  - name: Phase 1
    status: completed | in-progress  # Either indicates work done

Also check for completed tasks in plan.md or fix-plan.md:

• Search for - [x] (checked checkboxes) in the plan file
• Any checked task indicates implementation has started

For bugs, check .ai/bugs/{name}/fix-plan.md:

• Search for - [x] (checked checkboxes)
• Any checked task indicates implementation has started

Implementation detection summary:

Indicator	Location	Condition
Plan state status	plan-state.yml	status: in-progress or completed
Phase status	plan-state.yml	Any phases[].status: in-progress or completed
Completed tasks	plan.md or fix-plan.md	Any - [x] checkboxes found

Step 3: Determine prompt based on detection

If implementation has NOT started (no indicators found):

Default to plan verification mode. Display:

Found workflow: {name} ({workflow_type})
Status: Planning (no implementation started)

Proceeding with plan verification...

If implementation HAS started (any indicator found):

⚠️ STOP AND WAIT for user input. Display:

⚠️ Implementation in progress detected!

Workflow: {name} ({workflow_type})
Plan Status: {status from plan-state.yml or "Tasks started"}
Progress: {X} of {Y} phases completed | {N} tasks completed

Both the implementation plan AND code exist. What would you like to verify?

1. Verify implementation plan only
   - Check plan.md or fix-plan.md against coding standards
   - Useful if plan was recently updated

2. Verify code implementation (Recommended)
   - Check source files against plan and coding standards
   - You'll need to provide file paths

3. Verify both plan and code
   - Full verification of plan AND code against standards
   - Most comprehensive option

Please respond with 1, 2, or 3.

STOP AND WAIT for user response before proceeding.

Do NOT continue until user provides their choice (1, 2, or 3).

3. Read Artifacts to Verify

For Plan Verification Mode:

If workflow is a feature:

• Read .ai/features/{name}/implementation-plan/plan.md

• If missing:

  ⚠ Implementation plan not found for '{name}'.
  
  Run /ai.define-implementation-plan first.
  

If workflow is a bug:

• Read .ai/bugs/{name}/fix-plan.md

• If missing:

  ⚠ Fix plan not found for '{name}'.
  
  Run /ai.plan-fix first.
  

If workflow is an idea:

⚠ Ideas don't have implementation plans.

To verify an idea:
  1. Convert to feature: /ai.add "Feature based on {idea-name}"
  2. Create plan: /ai.define-implementation-plan
  3. Verify: /ai.verify

For Code Verification Mode:

Ask user for file paths:

Please provide the file paths to verify (one per line, or comma-separated):

Example:
  src/auth/login.ts
  src/auth/session.ts
  tests/auth.test.ts

Read the provided files.

For Both Mode (Plan + Code):

When user selects option 3 (verify both):

1. First, read the plan artifact:

   • For features: .ai/features/{name}/implementation-plan/plan.md
   • For bugs: .ai/bugs/{name}/fix-plan.md

2. Then, ask user for code file paths:

Plan loaded. Now provide the code file paths to verify (one per line, or comma-separated):

Example:
  src/auth/login.ts
  src/auth/session.ts
  tests/auth.test.ts

1. Read the provided code files.

2. Verification will check:

   • Plan against coding standards
   • Code against coding standards
   • Code