gco-2nd

GCO 2nd Opinion

Get a second opinion from GitHub Copilot CLI on a plan, approach, or codebase exploration.

Process

Step 0: Pre-flight Rate Limit Check

Before doing anything, check if Copilot is currently in degraded mode:

RATE_CHECK=$(node $HOME/.claude/scripts/gco-rate-limit.js check 2>&1)

If the output starts with degraded:, notify the user that Copilot is in low-cost mode (auto-downgraded model, free for Pro users) but proceed with Copilot anyway — it is still usable. Do NOT skip.

Step 1: Build Context Prompt

Construct a prompt that includes:

• The current plan or approach being considered
• Key files or areas of the codebase involved
• Specific questions or concerns

You are reviewing a development plan. Provide a second opinion.

## Context
<what the project is, what repo we're in>

## Current Plan
<the plan or approach being considered>

## Questions
1. Are there any risks, edge cases, or issues with this approach?
2. Is there a simpler or better alternative?
3. Are there any files or areas of the codebase that should be considered but aren't mentioned?
4. Any other suggestions or concerns?

Be concise and practical. Focus on actionable feedback. If the plan looks solid, say so briefly — don't invent problems.

Step 2: Run Copilot

LOGDIR=$(node $HOME/.claude/scripts/get-logdir.js)
mkdir -p "$LOGDIR"
DATETIME=$(date +%Y%m%d_%H%M%S)

bash $HOME/.claude/skills/gco/scripts/gco-run.sh \
  "<prompt>" \
  "$LOGDIR/${DATETIME}-gco-2nd.md" \
  "$LOGDIR/${DATETIME}-gco-2nd-stderr.log"

Timeout: 15 minutes.

Step 3: Collect Results

1. Check for quota fallback — grep the stderr log for GCO_USED_FALLBACK=:

   grep '^GCO_USED_FALLBACK=' "$LOGDIR/${DATETIME}-gco-2nd-stderr.log"
   

   If found, gco-run.sh auto-retried with gpt-4.1 because the primary model was out of quota. Notify the user with one line: "Used gpt-4.1 instead of claude-opus-4.6 because of no quota." Proceed — the feedback is still valid.

2. Read the output file

3. If empty or missing, check stderr log for errors

4. If Copilot timed out or produced no output, skip — the plan proceeds without the second opinion

Step 4: Return Feedback

Return the Copilot feedback to the caller. The caller decides whether to incorporate it into the plan.

No fallback needed. If Copilot fails, skip — the plan proceeds without the second opinion. This is advisory, not critical.

When Called by /x-as-pr or /x-wt-teams

1. After forming an initial plan, send it to Copilot for feedback
2. If useful feedback is returned, update the plan
3. Optionally re-run with the updated plan (up to 3 iterations total)
4. Finalize and proceed with implementation

Important Notes

• Copilot reads workspace files for context — no need to paste file contents into the prompt
• All file writing done by Claude Code, never by Copilot
• This is advisory — never block the workflow if Copilot is unresponsive
• Quota fallback policy: When the primary model returns HTTP 402 no-quota, gco-run.sh auto-retries with gpt-4.1 (free) and writes GCO_USED_FALLBACK=gpt-4.1 ... to the stderr file. Claude MUST check stderr for this marker and tell the user "Used gpt-4.1 instead of claude-opus-4.6 because of no quota." Feedback is still valid — only skip when Copilot actually fails (timeout, no output, not installed)
• Cheap variant: Use /gcoc-2nd to skip opus entirely and run gpt-4.1 from the start
• NEVER use ~ in paths — use $HOME