Nimi Coding

Nimi Coding provides deterministic access to project-owned canonical authority for AI hosts and repository tooling. It handles authority formatting, schema validation, relationship queries, dependency navigation, and impact review; it is designed as a focused specification tool rather than an autonomous AI agent, code generator, or product-spec synthesizer.

This documentation corresponds to @nimiplatform/nimi-coding@0.6.5 pinned by this workspace. Command ergonomics and parameters are described authoritatively in the package's CLI help and bundled README.

Specification and Configuration Boundaries

Path Pattern	Role & Scope
.nimi/spec/**/*.authority.{yaml,md}	The single canonical product authority source
.nimi/config/**	Repository-owned Nimi Coding host execution configuration
.nimi/methodology/authority-authoring.yaml	Official methodology guide for authoring authority
.nimi/local/**	Local verification artifacts and cache outputs (git-ignored; never authoritative)
config/**	Static inputs for apps, code generators, and host implementations; outside Nimi Coding configuration

Nimi Coding maintains a strict single source of truth: downstream developer documentation, test fixtures, compiler outputs, or audit reports must never override or redefine product authority.

Bounded Code Context Reads

• code context: Analyzes a specified TypeScript or TSX file, an admitted top-level symbol (exported or unexported), and tsconfig to extract a bounded outbound static context.
• code authority: Locates explicit @nimi-authority and @nimi-deprecated marker comments in Git-tracked TypeScript, TSX, Go, Python, and Rust source files, or inspects authority bindings for a given file.

Both commands are request-local and read-only. AI hosts invoke them on demand; they do not install mandatory preflight hooks, alter host task state, or attempt to evaluate unannotated code.

Authority Maintenance Workflow

1. Discovery: When the exact authority identifier is unknown, run authority discover within a bounded scope to locate candidate units.
2. Context resolution: Before making changes, retrieve the declared outbound context via authority context.
3. Format and validate: After editing, format each changed file with authority fmt and run authority check across the full .nimi/spec corpus.
4. Impact analysis: Run authority diff and authority impact with explicit byte budgets to review semantic blast radius.
5. Refactoring sweep: When planning architectural changes, use authority change-candidates to recall impacted nodes across deterministic channels. This output serves as recall input rather than automated arbitration.

The repository instructions in AGENTS.md and .nimi/methodology/authority-authoring.yaml define authoring guidelines in detail.

Next Steps

• Host Integration Guide
• CLI Reference

Source Basis

• nimi-coding/README.md
• .nimi/methodology/authority-authoring.yaml