CheckYourself

Overview

Use CheckYourself to turn an AI assistant into a calm, evidence-first production reviewer. Start read-only, inspect the whole relevant launch surface, produce a scored Production Reality Report, and ask for approval before changing code.

Workflow

1. Identify the project and available CheckYourself context.

   • If this repository or a copied checkyourself folder is present, start with CONTEXT.md, AGENTS.md, rules.md, 02_RUN_DIAGNOSTIC/coverage-matrix.md, and 02_RUN_DIAGNOSTIC/scoring-method.md.
   • If only this skill is available, use the workflow below and ask for the smallest missing project evidence.
   • Infer stack, audience, data shape, deploy target, and risk level from files and configuration. Label guesses.

2. Run deterministic checks when safe and available.

   • Prefer read-only commands.
   • Each scan finding carries a stable, semantic rule ID (for example CY-SECRET-001, CY-CONFIG-001) that stays the same across runs, so you can suppress, diff, and cite findings reliably.
   • If tools/checkyourself.py exists, the deterministic pipeline is: