Codex Analysis

Run the Codex CLI to get deep analysis using gpt-5.2-codex with high reasoning effort.

IMPORTANT: This skill MUST spawn subagents for parallel exploration. Always include subagent instructions in the prompt.

When to Use

• User asks for "Codex analysis" or "Codex help"
• User wants a second opinion on code or architecture
• User needs deep reasoning on a complex problem

Running the Analysis (MUST spawn subagents)

Execute this command, replacing {PROMPT} with the user's question or analysis request.

Always append subagent instructions to the prompt:

codex exec -m gpt-5.2-codex \
  --config model_reasoning_effort="high" \
  --sandbox read-only \
  --skip-git-repo-check \
  "{PROMPT} Spawn subagents to explore different aspects in parallel."

Command Parameters

Subagent Requirements

MANDATORY: Every prompt MUST include instructions to spawn subagents. This enables:

• Parallel exploration of different code areas
• Concurrent analysis of multiple concerns (security, performance, architecture)
• Faster and more thorough analysis

Template suffix to append to every prompt:

"Spawn subagents to explore different aspects in parallel."

After Running

1. Summarize the Codex analysis output
2. Highlight key suggestions and findings
3. Ask if the user wants to:
   • Implement any suggested changes
   • Get more details on specific points
   • Run additional analysis

Example Usage

User asks: "Use Codex to analyze the authentication flow"

Run (note: subagent spawning is REQUIRED):

codex exec -m gpt-5.2-codex \
  --config model_reasoning_effort="high" \
  --sandbox read-only \
  --skip-git-repo-check \
  "Analyze the authentication flow in this codebase. Spawn subagents to explore security issues, improvement opportunities, and best practices in parallel."

User asks: "Get Codex help with performance issues"

Run:

codex exec -m gpt-5.2-codex \
  --config model_reasoning_effort="high" \
  --sandbox read-only \
  --skip-git-repo-check \
  "Identify performance bottlenecks in this codebase. Spawn subagents to analyze database queries, API endpoints, and frontend rendering in parallel."

Then summarize findings and offer follow-up actions.