anthropic-instructions

Employee-level Anthropic ("ant") prompts for GitHub Copilot and open-source coding agents (claw-code, openclaude, opencode). Injects leaked internal rules for honest verification, autonomy, and minimal user intervention.