gpt-advisor

name	gpt-advisor
description	GPT as security advisor and architectural critic. Sends code or decisions to GPT for independent second opinion. Strong in security audits, code review, and finding logical flaws. Don't use for routine coding or simple tasks.
disable-model-invocation	true
model	sonnet
effort	medium
allowed-tools	Read, Grep, Glob, Bash
argument-hint	["topic-or-file"]

GPT Advisor — Security & Architecture Critic

Role

GPT acts as an INDEPENDENT advisor to the GODMODE orchestrator. NOT an executor. Does NOT write code. Gives OPINIONS and finds WEAKNESSES.

When to Consult GPT:

1. Security Review

Send code → ask GPT to find vulnerabilities. GPT is strong at: SQL injection, auth bypass, data exposure, race conditions.

Prompt template:

You are a senior security engineer. Review this code for vulnerabilities.
Focus on: OWASP Top 10, authentication flaws, data exposure, injection attacks.
For each issue: severity, file:line, exploit scenario, fix.
Code: [paste code]

2. Architecture Critique

Send architecture decision → ask GPT to find flaws.

Prompt template:

You are a senior software architect. Critique this architecture decision.
Find: scalability issues, single points of failure, over-engineering, missing components.
Decision: [paste decision]
Context: [paste context]

3. /dispute Participation

When /dispute is running, GPT provides independent opinion.

Prompt template:

You are an independent technical expert.
Topic: [topic]
Position A argues: [summary]
Position B argues: [summary]
Give your independent analysis. Where do you agree/disagree with each?
What did both miss?

Rules:

• Always include FULL context (GPT doesn't see our conversation)
• Specify output format expected
• Compare GPT's response with our own analysis
• GPT advises, orchestrator DECIDES
• Never blindly follow GPT — evaluate critically