Clarx

Scoring Non-JS Repositories

What full vs. partial analysis means, and how to get a trustworthy score on Python, Go, and other stacks.

Scoring non-JS repositories

Clarx scores any codebase — but not every rule runs on every stack. This page explains exactly what you get on a repository without JavaScript/TypeScript sources, so a partial analysis is declared, never silent.


Full vs. partial analysis

Most of the standard operates on the filesystem and guidance layers, which are language-agnostic:

  • All Discoverability rules (D1–D6)
  • All Operational Guidance rules (O1–O5)
  • Filesystem-based Boundary Clarity and Edit Safety rules
  • Line-count rules (C2, E1) — line counting covers .py, .go, .rs, .rb, .java, and more

Five rules require a resolvable JS/TS import graph and are reported as not evaluated on other stacks:

RuleWhat it needs
B1 — circular importsimport graph
C3 — import surface per fileimport graph
C4 — high fan-in documentationimport graph
C5 — import chain depthimport graph
C6 — boundary surface in entry filesJS/TS entry-file analysis

When these don't run, the CLI says so explicitly:

Not evaluated for this stack: B1, C3, C4, C5, C6 — no resolvable
JS/TS import graph. Guidance and filesystem rules still apply.

Not-evaluated rules never move the score in either direction — they are excluded, not counted as passing.


What this does to confidence

Without an import graph, confidence drops one level:

Manifest presentImport graphConfidence
yesyeshigh
yesnomedium
nonolow

A Python or Go repo with a manifest scores at medium confidence — a meaningful, comparable result for the pillars that ran.


How to raise confidence on a non-JS repo

  1. Add a clarx-manifest.json — this is the single highest-leverage step. Declare generated, verificationCommands, workspaces, and commonTasks. All of the Operational Guidance pillar and much of Discoverability become fully evaluable.
  2. Add a CLAUDE.md or AGENTS.md with verification commands and common-change locations.
  3. Keep the filesystem rules honest — root entry count, utility dumping grounds, file sizes, and route shadowing are all fully evaluated regardless of language.
npx @clarxai/cli init   # scaffolds a starter manifest
npx @clarxai/cli score .

Roadmap

Native import-graph support for Python (and later Go) is on the roadmap; it will move B1/C3/C4/C5 from not evaluated to fully scored on those stacks. Until then, the honest contract is: guidance-layer analysis everywhere, graph-layer analysis on JS/TS.