Bootstrap Commission

Use this commission with an AI Peer starting a repository that does not meaningfully exist yet — an initialized Git checkout and a purpose, nothing more. Everything it needs from Truth Machine is fetched from the published site.

The ladder is Tier 0 then Tier 1, nothing higher: packet machinery is for truth Git does not already govern, and a new repository has none yet. The Peer scaffolds; the operator supplies the truth that goes into the scaffolding. A front door full of placeholders is worse than no front door.

Stand up the Truth Machine scaffolding in this repository.

Preconditions — ask the responsible operator for any that are missing, and
do not proceed without them:

1. What this repository exists to keep true, in one sentence.
2. The one rule a collaborator must not break.
3. Which facts this repository owns, and which belong to named other
   systems.
4. The verification command that must pass before a handoff, if one exists
   yet.

Read first:

1. https://kinra.ai/docs/truth-machine/start/
2. https://kinra.ai/docs/truth-machine/kit/
3. https://kinra.ai/docs/truth-machine/profiles/ai-peer/

Fetch https://kinra.ai/docs/truth-machine/manifest.json once; download each
template below from its templates[] entry and verify the recorded sha256
before use.

Climb the ladder one rung at a time. Fill every placeholder with real
content from the preconditions; leave none. After each rung, show the
operator the file before writing the next.

1. Front door — templates/front-door/AGENTS.md. Identity, start-here order,
   authority boundary, structural rules, verification. Add the CLAUDE.md
   stub it describes.
2. Status document — templates/status-document/STATUS.md, with a status
   line from templates/status-lines/vocabulary.md. The first phase is
   honest: "scaffolding, nothing deployed" is a valid current state.
3. Typed current projection — templates/current-projection/current.toml,
   declaring STATUS.md as its first record. Validate it against
   https://kinra.ai/docs/truth-machine/schemas/projection.schema.json.
4. Governed read — templates/governed-read/state_reader.py, copied and
   owned. Run it; make it exit 0 against the projection you declared.
5. Declaration — templates/declaration/truth-machine.toml recording rungs
   1 through 4 as adopted with their paths, and every remaining pattern
   declined with a real reason ("nothing deployed yet" is a real reason).
   Copy templates/declaration/check_adoption.py and run it.
6. Wire the reads into the repository's front door: the status document,
   projection, and reader are the Start here list; the reader is the first
   command of any session.

Prove completion by reporting the observed exit codes of the governed read
and the adoption check, and the answers a cold session would give to the
six durable-office questions in the AI Peer profile — from the workspace
alone, no conversation memory. A question the workspace cannot answer is a
missing rung, not a footnote.

Evaluation questions

After the commission, ask the Peer these questions without additional context:

1. What does this repository exist to keep true?
2. What is current right now, and how is that answer produced?
3. What may a collaborator not do here?
4. What was declined in the declaration, and when should each decline be revisited?
5. What survives the loss of this Peer?

The answers should come from the scaffolded files, not from the session that wrote them.

Related

The Migration Commission covers a repository that already exists. The Adoption Guide is the judgment layer above both: whether the architecture fits at all, and which seam to pay for first.