Check command reference

CLI index

init_check — Check Project Against AXM Standard

Text Only

axm init_check [OPTIONS] [PATH]

Option	Short	Type	Default	Description
PATH		string	.	Directory to check
--json-output		bool	False	Output as JSON
--agent		bool	False	Compact agent-friendly output
--verbose		bool	False	Show all checks including passed
--category		string	all	Filter to one category

Default Python categories: pyproject, ci, tooling, docs, structure, deps, changelog, workspace, paper, experiment. The selected framework can change this registry; see check catalogue.

protocols is an explicit-only Python category. Run it with axm init_check PATH --category protocols; it is intentionally absent from an unfiltered run so projects outside the protocol profile keep identical scores and check counts. The check reads [tool.axm-init.protocols] and Python source statically through axm-ast; it never imports the inspected package.

Exit codes:

• 0 — Score is 100/100, or no weighted checks apply (N/A)
• 1 — Applicable score below 100, or an operational error

Example:

Bash

axm init_check --verbose

Text Only

📋 AXM Check — my-project
   Path: /path/to/my-project

  pyproject (29/29)
    ✅ pyproject.pyproject_exists        4/4  pyproject.toml found
    ...

  Score: 97/100 — Grade A 🏆

  📝 Failures (1):

  ❌ docs.readme (3 pts)
     Problem: README missing 1 section(s)
     Missing: Development
     Fix:     Add Development section(s) to README.md.

Check output with workspace context:

Bash

axm init_check

Text Only

📋 AXM Check — my-workspace
   Path: /path/to/my-workspace
   Context: WORKSPACE

  pyproject (29/29)
    ✅ pyproject.pyproject_exists        4/4  pyproject.toml found
    ...

  Score: 100/100 — Grade A 🏆

JSON output:

Bash

axm init_check --json-output

JSON

{
  "score": 97,
  "grade": "A",
  "context": "standalone",
  "workspace_root": null,
  "excluded_checks": [],
  "passed_count": 40,
  "failures": [
    {
      "name": "docs.readme",
      "message": "README missing 1 section(s)",
      "details": ["Development"],
      "fix": "Add Development section(s) to README.md."
    }
  ]
}

Output contract

The JSON example illustrates the schema; its score and counts are not a measurement of your project. --agent and --json-output both select the compact structured data shown above. The default is compact text, while --verbose selects the expanded human report. JSON/agent takes precedence over verbose.

A not-applicable run has score: null and grade: null; default text shows N/A. This is distinct from an applicable zero score. Treat N/A explicitly in CI rather than comparing null with a number. The tool writes a local quality snapshot as part of a check run.

Operational errors may have no structured score payload and still appear on stderr with a nonzero exit. The CLI does not expose an --framework override; framework detection happens from the target files.