promptkit
Installation
SKILL.md
promptkit
Two modes, split by artifact, because their rules contradict each other:
taskis the instruction you're about to hand an agent in this session. Ephemeral: one send, then it's dead. Context gets baked in, placeholders are forbidden, and it can lean on the repo.systemis the prompt your application sends on every request. Durable: its variables are the interface, and it has to survive input written by someone trying to break it.
What promptkit is not
- Not the task executor. Handed "add auth", it writes a prompt about adding auth. It does not add auth. This is the most likely failure, because the model will drift toward doing the task.
- Not the skill author. Authoring or improving a
SKILL.mdbelongs to a skill-authoring pass, meaning skillkit when installed. promptkit never writes one, and routes an ask to sharpen one. - Not the prose editor. A prose humanizer (humankit when installed) removes structure; promptkit adds it, because a machine is reading. Never run an em-dash rule or an AI-vocabulary list against a prompt.
- Not the project surveyor. promptkit reads one prompt. It never surveys the repo to decide what you should do next.
- Not an eval framework. Must-pass cases are a table you read. No harness, scoring code, metrics, A/B versioning, or token budgets.
- Never unattended. The deliverable is a prompt a human reads and sends.
When this fires
task is the default. State the mode you ran either way.