Truth Machine / Kit

Truth Machine Kit

Copyable, ownable patterns and templates for standing up a Truth Machine without reading five repositories.

Read as Markdown

The Kit is the layer between the normative specification and the implementation evidence: the reusable pieces named as pieces, each with its cost, prerequisites, and a copyable starting file where one helps.

It is a catalog, not a framework. A repository adopts a pattern by copying and owning it; nothing here is imported, versioned against, or depended on after adoption. Every page states its normative weight explicitly: either it implements identified TM-* requirements, or it is ergonomics — conventions that make the requirements cheap to live with. Roughly half the catalog is ergonomics, and a project can skip those without guilt.

Composition rules

Tiers group the patterns by what they cost and what must already exist. They are à la carte with stated prerequisites, not a strict ladder — running Tier 1 without Tier 2 is a valid configuration, and one evidenced implementation does exactly that.

Git already governs source code. Tier 3 packet machinery exists for truth Git does not already version and review — deployments, organizational facts, financial records, outside coordination. In an ordinary software repository, evidence and change packets duplicate what commits and PR review already do. What Git does not govern there is what is actually deployed, what was decided and why, and whether a document is still current — Tiers 0–2.

The tiers

# Pattern Tier Normative weight
1 Front Door 0 ergonomics, supports TM-DEC-006
2 Status Lines 0 ergonomics
3 Status Document 0 ergonomics
4 Authority Boundary 0 implements TM-MOD-005, supports TM-CUR-008, TM-CUR-009
5 Evidence Hygiene 0 implements TM-EVD-008
6 Frozen Archive 0 supports TM-CUR-006
7 Decision Records 0 ergonomics
8 Typed Current Projection 1 implements TM-CUR-001, supports TM-MOD-004
9 Human Index 1 supports TM-OBS-004
10 Governed Read 1 implements TM-OBS-001TM-OBS-003
11 Sealed Baseline 1 supports TM-CUR-006, TM-CUR-009
12 Declared Versus Observed 2 implements TM-OBS-005TM-OBS-009
13 Untrusted Intake 3 implements TM-EVD-001, supports TM-PUB-006
14 Evidence Packet 3 implements TM-EVD-002, TM-EVD-003, TM-EVD-007
15 Change Packet 3 implements TM-REC-004TM-REC-008, TM-CHG-009
16 Atomic Transition 3 implements TM-CHG-003TM-CHG-008
17 Exact Review 3 implements TM-CHG-001
18 Publication Freeze 4 implements TM-PUB-001TM-PUB-004
19 Generated Artifacts 4 supports TM-PUB-002, TM-OBS-003

At the Tier 4 boundary, Facets is the established pattern for projections that meet a person outside the operating boundary; it predates the Kit and keeps its own page.

Reading a pattern page

Every page uses the same shape. Normative weight is the load-bearing field: it reads either ergonomics or implements followed by identified requirements, meaning the pattern is one concrete way to implement them — the verb is implements, never satisfies, because a catalog page cannot grant conformance to anyone (TM-CON-003 reserves satisfied for evidence-backed dispositions inside an assessment). Evidence comes in two classes: assessed links an exact-revision implementation report; observed names a repository the maintainer operates without a pin and carries no conformance claim.

Declaring adoption

A repository can record which patterns it adopted in a machine-readable declaration: truth-machine.toml, validated by declaration.schema.json. Declining a pattern requires a stated reason — deliberate omission stays cheap; silent omission becomes impossible.

The companion check_adoption.py verifies the declaration against the working tree: every adopted pattern names paths that exist, every declined pattern states its reason. A finished migration or bootstrap is then an exit code, not a judgment; its --json output shares the governed read’s read-result.schema.json contract.

Alignment with open conventions

The Front Door pattern aligns with AGENTS.md, the open agent-guidance format launched in August 2025 and donated to the Linux Foundation’s Agentic AI Foundation in December 2025. The Kit extends that industry convention; it does not invent a rival one.