Use case · Build faster

Resolve code and policy drift.

Put your architecture, standards, and decisions where the agent reads them and CI enforces them — instead of in a rules file that grows until it stops working.

What happens today

  • A rules file per tool — CLAUDE.md, AGENTS.md, .cursorrules — drifting from each other and from the code.
  • Specs that govern generation and then die at merge, with nothing reconciling them against what shipped.
  • Structural lint that checks who imports whom, but never whether this is the approved calculation.
  • AI code review as the new bottleneck: same diff, different findings, and agentic pull requests waiting far longer for a reviewer.

Every layer is either prose without enforcement or enforcement without semantics.

How it works

A small set of SIGN™ documents lives in every repository: what the module is, its architecture, its dependencies, its public surfaces, its standards, its decisions. Human-readable, versioned, and machine-readable. Agents read them as governing context. CI compiles them as a gate. What can be generated from code is generated from code, so those parts cannot drift.

Explainer

Code

A coding agent drifts from the standard in a single session, then works from the governed contract instead.

Explainer

What is SIGN™

Prose an agent has to interpret, next to the same rule expressed as SIGN™.

Proof

We ran the gap analysis against our own estate: roughly 50 legacy repositories scanned against governed business rules, returning 412 identified rules with specific line-level conflicts. Our engineering baseline was 15–20 story points per sprint; on Canon we put 174 story points into production in four days.