README.md

Documentation

Start with the tutorials in order. Reach for a guide when you have one job to do. Read a concept page when you want to know why. Use the reference for every detail.

Tutorials — learn by doing, in order

#	Tutorial	You end with
1	Adopt your first frompt	an agent that adopted a frompt, said so, and let it go on command
2	Write a frompt	a frompt that lints, in the layout every tool reads
3	Publish a catalog	a signed catalog anyone can verify against your key
4	Run unattended	agents resolving by id from a catalog registered once, refusing what does not verify

Guides — one job each

Guide	When you want to
Talk to a ghost	address an adopted frompt, your agent, or relay between them
Revoke a version	withdraw a published frompt that is wrong
Attest with assay	record what a scanner saw about each document, without letting it become a verdict
Keep a catalog fresh	renew a manifest before clients refuse it
Run the conformance harness	see what real agents actually do with frompts
Use the Claude Code plugin	find, adopt and list frompts from inside Claude Code

Concepts — why it is shaped this way

Page	Explains
Consent, not security	what adoption settles, and every way an adopted frompt can still go wrong
Late binding	why a skill should point at a frompt instead of containing it; versions
What a frompt can be	a method, a rubric, a colleague, a game, a front door

Reference

Document	Covers
FPA.md	the protocol, normative
CLIENT.md	what the software around an agent must do, in four levels
SECURITY.md	the trust model, and what to look for when you read a frompt
TERMINOLOGY.md	the vocabulary, and the words that collide
INDEX.md	every frompt in this repository, with digests
Tools	every fp-* tool by role, with usage and exit codes
Repository layout	where everything lives

Examples

../examples/, smallest to full: one adoption, two and a disown, refusals, a recorded game, a recorded ghost interview, and a catalog end to end that make test runs.