wtf is dictator-skill?

Dictator is a shareable Agent Skill that gives blunt, ownership-level feedback on a product or project. Instead of politely working within the assumptions you already made, it questions the direction itself and pushes back on the plan as a whole. Running it produces a verdict on the project, a list of things to kill, a suggested strategic bet, a list of things worth keeping, and a priority order for what to tackle next. It is read-only and advisory: it does not change any files in your project, it only gives an opinion. The project is built around the Agent Skills format, a shared open layout that multiple AI coding agents can read, rather than a plugin format tied to one specific provider. This means the same skill file works across different agents instead of being locked to just one tool. Installing it is done through a command line tool called the Skills CLI, using an npx command that detects which AI agents you have installed and asks you to confirm where to install the skill. It can be installed globally so it is available in every project, or locally so it only applies to the current project and can be shared with teammates through the repository itself. If the CLI tool is not available, the skill files can also be copied by hand into the folder your agent expects. Once installed, you can trigger it by asking your agent to use dictator mode, or with more casual phrases like asking it to be brutal or to challenge your direction. Some agents may also expose it as a slash command. The repository itself is written in Python for its small helper scripts that validate and package the skill, though using the skill day to day does not require Python. It is released under the MIT license and is an independent community project, not affiliated with or endorsed by Anthropic.