Clauding Manifesto

meta

How to set up your project for successful agentic engineering.

Short rules that keep a repo's knowledge agent-usable and its "done" provable. Strong models follow short rules well: text gets most of the way, a small checker covers the rest.

rule	lives in	principle
QA	`QA.md` at the project root	evidence, not a green checkmark
current architecture	`docs/architecture.md`	architecture
decisions	`docs/adr/NNNN-slug.md`	ADRs
papercuts	`docs/papercuts/slug.md`	papercuts
writing style	`docs/style.md`	style
small linked tagged docs	`docs/**` + `docs/tags.md`	knowledge
freshness	frontmatter `code:` refs	freshness

Use it

situation	do
new repo	copy the scaffold, paste the rules into `AGENTS.md`
existing repo	paste the catch-up prompt into an agent session at the repo root
repo on an older version	same prompt: it reads the `project-knowledge rules vN` marker and upgrades
agent reading this site	start at `llms.txt`

Evolving

Practices on trial, not in the rules yet: session handoff, ideas inbox. All.

Scaffold

Copied into target repos as-is:

QA.md: tiers + expect in frontmatter, oracles table
docs/README.md: index
docs/architecture.md: diagram, component table, invariants
docs/style.md: writing rules
docs/tags.md: seed vocabulary
docs/adr/0000-template.md
docs/papercuts/_template.md

This site follows its own rules

Its architecture, decisions and papercuts are published alongside. Source: github.com/mojzis/clauding.

raw markdown