AI Coding Checklist is a skill published by agensi that runs a structured pre-task audit before an AI coding agent begins work. It addresses a specific failure mode: agents producing incorrect or harmful code because the task definition was too vague. Before code generation starts, the skill evaluates five gates — requirements, scope, context, testing strategy, and blast radius — and issues a verdict of either GO or NOT READY. If the verdict is NOT READY, the skill produces a BLOCKERS list that identifies the exact questions that must be answered or the specific files the agent must read before it is cleared to proceed. The output is a structured report with a visual verdict, giving developers a concrete record of what was verified and what still needs clarification. The skill is designed to work with IDE agents such as Cursor, VS Code, and GitHub Copilot, CLI agents such as Claude Code and Aider, and research and development environments such as ChatGPT. It does not write code itself; its sole function is to verify that the preconditions for a coding task are sufficiently defined so that the agent can act within safe, well-scoped boundaries. Use it at the start of any non-trivial coding task where incomplete context could cause the agent to break existing systems or miss acceptance criteria.

Use cases

Run a pre-flight audit before asking an agent to refactor a large module

Verify scope and blast radius before an agent modifies shared infrastructure code

Identify missing acceptance criteria before an agent writes a new feature

Check that a testing strategy is defined before an agent generates test suites

Surface missing file dependencies an agent needs to read before starting work

Gate a coding task to ensure requirements are complete before consuming tokens

When to use it

Before any AI coding task where requirements or scope are incompletely defined

When working with agents in Cursor, VS Code, GitHub Copilot, Claude Code, or Aider

When changes could affect shared or critical systems and blast radius needs to be explicit

When past agent runs have gone off-rails due to vague instructions

When not to use it

When the task is trivial and requirements are already fully specified

When no AI coding agent is involved — this skill is specifically a pre-agent coordinator

When you need the skill to generate or execute code directly; it only audits and reports

When your agent environment is not one of the listed supported tools, as compatibility is unconfirmed

Ai Coding Checklist

Use cases

When to use it

When not to use it