---

name: Deep Code Audit description: Perform a comprehensive, cognitive static analysis of a specific feature or logic flow to identify logical flaws, missing connections, or design inconsistencies.

Deep Code Audit Protocol

Use this skill when the user asks to "audit", "check logic", "verify flow", or "deep check" a specific feature, or when you are debugging a complex issue where automated tests are passing but behavior is wrong.

1. Define Scope

Clearly identify the Single Source of Truth for the feature being audited.

• Entry Point: Where does the logic start? (e.g., MainScene.update, EventBus.on('START_MATCH'))
• Key Components: Which Managers, Systems, or Classes are involved?

2. Gather Context (Crucial)

You MUST read the current state of the files. Do not rely on memory.

• Use view_file on all relevant files identified in Step 1.
• Look for imports to see what dependencies are brought in.

3. Execution Trace (The "Dry Run")

Mentally simulate the code execution line-by-line for the critical path. Checklist for Tracing:

• Data Units: Are time units consistent? (ms vs seconds). Are angles consistent? (Radians vs Degrees).
• Null Safety: Is this.manager or this.player checked before access?
• State Sync: If A updates a value, does B read the new value or the old one? (Order of operations).
• Event Linkage: If an event is emitted (emit), is there an active listener (on)? Is the listener cleaned up (off)?
• Boundary Conditions: What happens at 0 HP? What happens at Wave 0? What happens if the array is empty?

4. Verify Against "SKILL: Delivery Sentinel"

• Deep Code Audit finds logic errors.
• Sentinel finds syntax/runtime errors.
• If you find a logic error here, consider suggesting a new test case for the Sentinel to prevent future regression.

5. Report Format

Present your findings in a structured "Audit Report":

🔍 Deep Code Audit: [Feature Name]

1. Execution Flow

Briefly describe the path: Input -> Process -> Output.

2. Status Checks

• Data Integrity: ✅ Pass / ❌ Fail (Explain)
• Null Safety: ✅ Pass / ❌ Fail (Explain)
• Logic Consistency: ✅ Pass / ❌ Fail (Explain)

3. Critical Findings

• List any blocking issues or bugs found.

4. Recommendations

• Refactoring suggestions or missing safeguards.