gsd:second-opinion

GSD Second Opinion

Query OpenAI's Codex CLI for an independent perspective on code, architecture, or implementation decisions.

Usage

/gsd:second-opinion [your question]

Examples

/gsd:second-opinion Review the auth flow for security issues
/gsd:second-opinion Is there a better way to structure the API routes?
/gsd:second-opinion Check if the database schema handles edge cases
/gsd:second-opinion What did I miss in this implementation?

How It Works

1. Runs codex-oracle with gpt-5.2-codex model and xhigh reasoning
2. Codex sees the same codebase (read-only access)
3. Returns Codex's analysis without the full execution trace

Execution

Run the codex-oracle script with the user's question:

QUESTION="$ARGUMENTS"
if [ -z "$QUESTION" ]; then
  echo "Usage: /gsd:second-opinion [your question]"
  echo ""
  echo "Examples:"
  echo "  /gsd:second-opinion Review the error handling in src/api/"
  echo "  /gsd:second-opinion Is this the right architecture for real-time updates?"
  exit 1
fi

# Support both project-local (.claude/) and global (~/.claude/) installs
if [[ -f ".claude/skills/codex-oracle/scripts/ask_codex.sh" ]]; then
  RESULT=$(bash .claude/skills/codex-oracle/scripts/ask_codex.sh "$QUESTION")
else
  RESULT=$(bash ~/.claude/skills/codex-oracle/scripts/ask_codex.sh "$QUESTION")
fi
echo "$RESULT"

When to Use

• Before major decisions: Get Codex's take on architectural choices
• After implementation: Independent code review
• When stuck: Different perspective on debugging
• For verification: Confirm Claude's analysis is complete

Integration with GSD

This skill is also called automatically by GSD workflows:

• verify-phase.md - Parallel verification + final check
• diagnose-issues.md - Alternative hypotheses
• execute-plan.md - Post-completion code review