Are minor code fixes allowed without blocking?

Yes, syntax corrections, variable naming, and internal code organization are allowed as long as they do not change the underlying approach or architecture.

What is the Following-Plans skill used for?

It is used to ensure that AI agents follow a pre-defined implementation plan exactly, preventing them from making unauthorized architectural or logic changes without human approval.

When does an agent report a BLOCKED status?

An agent reports BLOCKED whenever a task requires a different algorithm, library, data structure, or API design than what was specified in the original plan.

How does this skill prevent scope creep?

It includes a 'Rationalization Defense' that identifies common excuses agents use to deviate from a plan, such as finding a 'simpler way,' and forces a stop-and-report workflow instead.

Can I use this skill for manual coding?

While designed for AI agents, the decision tree provides an excellent framework for human developers to ensure they stay aligned with technical specifications during implementation.

Plan Execution Guard

Name: Plan Execution Guard
Author: cipherstash

bycipherstash

•

생산성 및 워크플로

Enforces strict adherence to implementation plans by providing an algorithmic decision tree to prevent unauthorized architectural deviations and scope creep.

The Following-Plans skill provides a critical governance layer for AI agents during the execution of software development tasks. By utilizing a standardized decision tree, it forces agents to distinguish between minor syntax corrections and significant architectural shifts. This prevents agents from making silent 'simpler' changes that bypass design decisions, ensuring that all deviations are explicitly reported and approved. It is an essential tool for maintaining technical integrity and preventing unauthorized refactoring during the implementation phase.

주요 기능

01Standardized OK/BLOCKED status reporting framework

02Clear distinction between syntax fixes and approach changes

03Algorithmic decision tree for implementation governance

04Automated identification of unauthorized architectural changes

05Rationalization defense against AI-driven scope creep

064 GitHub stars

사용 사례

01Maintaining architectural integrity during complex feature implementations

02Governing AI agents to prevent silent refactoring or library changes

03Standardizing feedback loops between planning and execution phases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cipherstash/cipherpowers following-plans

For use in Claude.ai and ChatGPT

Download Skill