Skill

gco

Runs GitHub Copilot CLI in read-only mode as sub-agent for code review, research, and second opinions on code or architecture. Invoke via /gco or auto-activates on 'gco' or 'copilot' mentions.

Bash

Node

Github

code-quality

developer-tools

npx claudepluginhub takazudo/claude-resources

Configuration

Arguments: <prompt>

Tool Access

This skill is limited to using the following tools:

Bash(bash *)Bash(timeout *)Bash(gtimeout *)Bash(node *)

Preview

Run GitHub Copilot CLI in read-only mode for code review, research, and investigation tasks.

Supporting Assets

scripts/gco-pr-body.shscripts/gco-pure.shscripts/gco-run.shscripts/gcom-msg.sh

SKILL.md

Similar Skills

cache-components

139.3k

Guides Next.js Cache Components and Partial Prerendering (PPR): 'use cache' directives, cacheLife(), cacheTag(), revalidateTag() for caching, invalidation, static/dynamic optimization. Auto-activates on cacheComponents: true.

cache-components

mcp-builder

124.2k

Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).

9 files

anthropics-skills-13

Stats

Stars7

Forks0

Last CommitMay 7, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

GitHub Copilot Read-Only Sub-Agent

Run GitHub Copilot CLI in read-only mode for code review, research, and investigation tasks.

How It Works

Copilot runs non-interactively with -p (prompt) and -s (silent output)
Read-only enforced via --deny-tool='write' (hard tool-level block) and read-only preamble in prompt
Uses --no-ask-user and --allow-all-tools for fully autonomous execution
Model: claude-opus-4.6 (primary), auto-fallback to gpt-4.1 (free tier) on no-quota
Timeout: 15 minutes

Cheap variant

If you already know Premium quota is low and want to skip the opus attempt, use /gcoc instead — identical behavior but forces the free gpt-4.1 model from the start.

Process

Step 1: Build the Prompt

Take the user's request and use it as the copilot prompt. If the user provides /gco with arguments, use those as the prompt directly.

Step 2: Run Copilot

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

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

Run as a background Bash task with 15-minute timeout.

Step 3: Collect and Present Results

Read the output file ($LOGDIR/${DATETIME}-gco-${SLUG}.md)
If empty or missing, check stderr log for errors
Check the stderr log for a GCO_USED_FALLBACK= line. If present, 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 with the output — it is still valid.
Synthesize and present findings to the user
Include the log file path for reference

Important Notes

Copilot cannot modify files — all writes are done by Claude Code
NEVER use ~ in paths — use $HOME
Output and stderr go to $LOGDIR/${DATETIME}-gco-* (timestamped)
If copilot is not installed, report "Copilot CLI not found. Install with: gh extension install github/gh-copilot" and fall back to Claude Code tools