How does the risk tier system work?

Plans are classified from Tier S (safety-critical) to Tier C (trivial). Higher tiers activate more rigorous audit gates, ensuring complex tasks get more scrutiny.

When should I use the Plan Audit skill?

It should be used immediately before finalizing any implementation plan or exiting 'Plan Mode' to ensure the proposed steps are safe and technically sound.

What is the Killer 7 universal core?

The Killer 7 is a set of non-negotiable audit questions covering premortems, math validation, source evidence, and blast radius analysis that apply to every plan.

What are 'AI-isms' in a plan audit?

AI-isms are patterns where an AI provides overly generic or non-technical solutions. The audit identifies these to ensure the plan meets senior engineering standards.

Plan Audit & Risk Guard

Name: Plan Audit & Risk Guard
Author: Undercurrent-Studio

byUndercurrent-Studio

0•

보안 및 테스팅

Performs a rigorous 44-gate audit of implementation plans to catch architectural flaws, security gaps, and mathematical errors before code is written.

Plan Audit serves as a mandatory quality assurance layer for AI-driven development, preventing silent failures and technical debt. By enforcing the 'Killer 7' universal core—which includes arithmetic validation, premortem analysis, and evidence-based data checking—this skill ensures every proposed change is viable. It utilizes a risk-tiered depth system (S through C) to apply the appropriate level of scrutiny to everything from trivial UI tweaks to safety-critical database migrations and API overhauls, forcing the AI to prove its logic with data rather than assumptions.

주요 기능

01Killer 7 universal audit core for every implementation plan

02AI-ism smell test to eliminate generic or hallucinated logic

03Arithmetic-based resource modeling and math validation

04Domain-specific gate activation for databases, APIs, and pipelines

05Risk-tiered depth controls ranging from trivial to safety-critical

060 GitHub stars

사용 사례

01Auditing high-stakes security or authentication logic before implementation

02Calculating resource constraints and API rate limits for batch processing tasks

03Verifying complex database migrations and schema changes for data integrity

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add undercurrent-studio/undercurrent-cortex plan-audit

For use in Claude.ai and ChatGPT

주요 기능

01Killer 7 universal audit core for every implementation plan

02AI-ism smell test to eliminate generic or hallucinated logic

03Arithmetic-based resource modeling and math validation

04Domain-specific gate activation for databases, APIs, and pipelines

05Risk-tiered depth controls ranging from trivial to safety-critical

060 GitHub stars

사용 사례

01Auditing high-stakes security or authentication logic before implementation

02Calculating resource constraints and API rate limits for batch processing tasks

03Verifying complex database migrations and schema changes for data integrity

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add undercurrent-studio/undercurrent-cortex plan-audit

For use in Claude.ai and ChatGPT