This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

Русский

Setup

How to bring codeharness into a new project. This is a working procedure, not a normative rule — codeharness itself doesn't (yet) prescribe an attach mechanism; see the note at the bottom.

What you're copying

From the codeharness repository into your target project:

• CLAUDE.md — the core file: router table + universal rules.
• .ai-workflow/*.md — the topic files the router table points at.

Both are meant to be adapted, not dropped in verbatim. codeharness's own README.md says it plainly: it is a source of templates, and every consuming project tunes it individually.

Procedure

1. Copy CLAUDE.md and .ai-workflow/ into the target project's root (or wherever your tooling expects project instructions — check your Claude Code setup).
2. Trim what doesn't apply. Not every project needs every topic file — e.g. a project with no formal spec doesn't need spec-discipline.md wired in (that file is explicitly conditional — see File Map). Remove the corresponding router-table row too, so the table doesn't point at a file you deleted.
3. Adapt what's project-specific.
   • issue-tracker.md assumes a Forgejo/Gitea/GitHub-style tracker (#N, closes #N, labels). A project on Jira or similar needs this file rewritten — the source file says so explicitly, it is not universal.
   • Branch roles (working/release/default) must be documented explicitly for the project — codeharness does not infer them from branch names.
   • The status/decisions file paths (.ai-workflow/memory/STATUS.md, .ai-workflow/memory/DECISIONS.md) are defaults; a project can point elsewhere as long as it says so.
4. Fill in what's missing. If the project wants routing enforced by a hook rather than convention, task-routing.md includes an optional UserPromptSubmit/PreToolUse hook template — copy and adapt, it is opt-in.
5. Record the codeharness origin, if useful for future maintenance — e.g. a line in the project's own README noting it derives its CLAUDE.md/.ai-workflow from codeharness and at roughly which version, so future updates can be diffed against upstream.
6. Verify the language state. As of this writing codeharness's CLAUDE.md/.ai-workflow are Russian-only (temporary, see documentation.md). If your project needs English instructions, that is a deliberate translation pass on your side, not a mechanical copy.

What codeharness does not yet give you

• No symlink/submodule/package mechanism. codeharness is not symlinked or copied into ~/.claude automatically, and there is no established update-propagation mechanism (e.g. "pull the latest git-discipline.md into 6 projects at once"). This is called out directly in codeharness's own README: "the mechanism for attaching to target projects is not yet defined, still under discussion."
• No scaffolding command. There is currently no generator/init script that does steps 1–4 above for you. Setup is manual copy + edit.

If you're setting up multiple projects and find yourself repeating this procedure often, that repetition is itself a signal for the attach-mechanism discussion upstream — consider raising it as an issue in codeharness rather than solving it silently per-project.

After setup

Once CLAUDE.md/.ai-workflow are in place, the project follows the cycle described in Process: plan → consent → issue → work → verification → review, with SemVer applied to the project's own VERSION, independent of codeharness's version.