Skip to content

Feature Map

Index combines human docs with machine-readable structure so repository intelligence can survive handoffs, audits, and agent sessions.

For the end-to-end flow, see How It Works.

Core Surfaces

SurfaceWhat it doesStart here
Workspace generatorCreates a complete index-style repository from explicit CLI flags or a config file.generator/create-index.mjs
ProfilesSelect default feature bundles for public templates, company brains, client workspaces, repo intelligence, and automation workspaces.generator/profiles/
Feature templatesEmit docs, repo registries, workspace records, automation runbooks, AI evals, MCP folders, and company-brain guidance.generator/templates/
Repository registryTracks codebases with owner, stack, purpose, status, tags, and start-here files.repos/index.yaml
Workspace metadataCaptures structured context for clients, projects, repositories, vendors, contacts, communications, opportunities, next actions, and evidence.schemas/workspace.schema.json
Validation toolsCheck generated workspaces, workspace metadata, people records, references, dates, enums, and suspicious secret-like key names.scripts/validate-workspaces.mjs
MCP serverExposes docs and repository registry entries as resources, plus search tools for docs and tracked repos.mcp/src/
Prompt libraryProvides reusable prompts for setup, investigation, rapid build, pilot rescue, and handoff work.prompts/
Project hygieneIncludes issue templates, a pull request template, generator tests, workspace validation tests, and Makefile checks..github/, generator/*.test.mjs, scripts/*.test.mjs, Makefile

Generated Feature Modules

Generated workspaces can include these modules:

  • docs: a documentation site with build, preview, validation, and LLM-readable output.
  • repo-registry: repos/index.yaml for tracked repositories and source systems.
  • workspaces: structured workspace records plus schema and validation scripts.
  • automation-runbooks: recurring automation instructions in repo-local docs and .codex/automations/.
  • ai-evals: smoke tests and fixtures for prompt, retrieval, tool, and output-format changes.
  • company-brain: operating model, roadmap, intake prompt, and digital-brain map template.
  • mcp: read-only MCP starter files for exposing approved context to agents.

How The Pieces Work Together

Built by Baker Street for evidence-backed company memory.