skills.rest
sanjanb
sanjanbCOMMUNITY
Explore all skills by this author →
GitHub

claude

Invokes Claude Code CLI for independent diff review, adversarial challenge, and read-only repo consultation.

1|Updated Jul 23, 2026
Search Tags:#code review#ai orchestration#claude cli#second opinion#diff review#adversarial review#gstack
One-click install
npx skills add https://github.com/sanjanb/my-agent-harness --skill claude-sanjanb
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: claude
Source: https://github.com/sanjanb/my-agent-harness/tree/main/skills/gstack-claude
Command: npx skills add https://github.com/sanjanb/my-agent-harness --skill claude-sanjanb

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When working in a non-Claude AI host, you lose access to Claude's perspective for code review and second opinions. This Skill wraps the Claude Code CLI so any host can request an independent diff review, an adversarial failure-mode challenge, or a read-only consultation about the repository. ## Core Features & Use Cases - Review mode: Runs an independent diff review via claude -p, giving you a second set of eyes on uncommitted changes before shipping. - Challenge mode: Performs an adversarial failure-mode review that hunts for edge cases and weaknesses in your implementation. - Consult mode: Asks Claude questions about the repo using read-only file tools, useful for architecture questions or onboarding to unfamiliar code. - Use Case: Before merging a branch, ask for a "claude review" to get an outside voice on your diff, then run "claude challenge" to surface failure modes your primary agent missed. ## Quick Start Ask the agent to run a claude review on my current diff to get an independent second opinion before I ship.

Frequently Asked Questions about claude

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get a second opinion on my code diff from Claude?▼

Invoke the skill with a phrase like "claude review" or "second opinion from claude". It runs an independent diff review through the Claude Code CLI in print mode, returning feedback on your uncommitted changes without modifying files.

What is the difference between claude review, challenge, and consult modes?▼

Review mode performs an independent diff review of your changes. Challenge mode runs an adversarial failure-mode review to find edge cases and weaknesses. Consult mode answers questions about the repo using read-only file tools.

Can I use Claude Code CLI from a non-Claude AI host?▼

Yes, this skill is designed exactly for that scenario. It wraps the claude CLI so hosts like opencode or other agents can invoke Claude for review, challenge, or consult tasks through shell commands.

Does the claude consult mode modify my repository files?▼

No, consult mode uses read-only file tools only. Claude can inspect and answer questions about your codebase but cannot write or edit files during a consultation.

Why does the skill run a long shell preamble before starting?▼

The preamble detects the gstack installation root, session kind, telemetry settings, and repo mode so the skill behaves correctly across interactive, headless, and spawned sessions. It is standard gstack setup, not specific to the Claude invocation.

Related Skills

Explore More Software Engineering Skills→
fwojciecfwojciec

gemini-review

Automate code reviews by querying Gemini on Git branch diffs.

Community
Advanced
aryaxtaryaxt

doit

Automate the complete pull request shipping workflow from change analysis to merge.

Community
Advanced
phraphra

code-review

Review modified code files and auto-fix mechanical issues.

Community
Advanced
tsenseitsensei
3

codex

Review code diffs with Codex CLI and log pass/fail gates.

Community
Advanced