Renma Documentation

Start with the README for an overview, then choose the reference for your task. Each linked guide owns its detailed contract; this index maps where to look.

Start Here

• README: product purpose, boundaries, installation, and first commands.
• User Manual: CLI workflows, operational metadata, and repository configuration. Emitted renma --help and renma <command> --help remain the command-line authority.
• Authoring Guide: asset responsibilities, evidence, creation gate, and handoff. Interaction guidance is adaptive, not a prescribed LLM reasoning algorithm.

Diagnostics And Security

• Diagnostics Reference: Finding semantics, stable IDs, evidence, repair constraints, and verification.
• Security Policy Guide: effective policy, revision review, and supported instruction analysis. The User Manual owns field mappings; the guide explains the boundaries with SAST, secret/dependency scanning, and runtime controls.
• Renma Quality Profile: exact thresholds, units, provenance, rationale, and configuration status.
• Metadata Budget: metadata-size diagnostics.
• Context Lifecycle Diagnostics, Context Conflict Diagnostics, and Context Language Diagnostics: focused diagnostic families.

Asset And Relationship Contracts

• Version-pinned Asset References and Builder API: exact references, version binding, and the read-only builder API.
• Agent Skills Compatibility and Migration: canonical paths and metadata, validation, and one-way migration.
• Advanced Skill Authoring: focused workflows, orchestration, and progressive disclosure.
• Context Lens: justification, fields, relationships, and the non-runtime boundary.
• Declared Composition: forward required/optional closure, provenance, completeness, conflicts, and cycles.
• Declared Impact: reverse explicit composition and change review.
• Skill Discovery: static publication, continuation routes, reachability, coverage, Skill Index, diff, and warn-only CI policy.
• Repository Context BOM v3: deterministic, reproducible declared manifests.
• Machine-Readable JSON Compatibility: top-level schemas, 1.x compatibility rules, and environment-derived fields.
• Experimental Execution Contract: static possible relationships, one-snapshot guarantees, bounded completeness, and external revision/hash binding.
• Trust Graph v2: evidence provenance and ordering, not a trust score.
• Published JSON Schemas: packaged BOM, Trust Graph, and Skill Authoring Handoff contracts.

Development

Files under development/ explain implementation rationale and are excluded from the npm package. User-facing contracts remain in the references above.

• Public Architecture: layers, data flow, and stable product boundaries.
• Internal Architecture: module ownership, dependencies, projections, compatibility seams, and invariants.
• Product Design: durable decisions and capabilities outside core.
• External Review Governance: candidate direction and SkillSpector experiment plan, not a current product contract.
• Current Roadmap: stabilization baseline, open candidates, deferred concerns, and non-commitments.
• Renma 1.0 Stabilization Boundary: completed reduction, retained shell/CI contracts, precision-first analysis, and release gates.
• Skill Discovery Design: rationale behind the current static Discovery contract.
• Versioned Asset Bindings Design: reference and metadata decisions for external builders.
• Release Publication Security: repository checks, their limits, and required external npm/GitHub controls.
• Changelog: release history and compatibility notes.

Examples

Examples require a source checkout and are not included in the npm package:

• Interactive Placeholder
• Example Context Repository
• Context Lens
• GitHub Actions

Maintaining The Documentation Site

Markdown in this directory is the documentation source of truth. From the repository root:

npm run docs:dev
npm run docs:build
npm run docs:preview

Fenced mermaid blocks render automatically; their source remains the editable form. Verify diagram changes with npm run docs:build and a local preview. Diagrams must not require loose security or executable click behavior.