# AI use cases for Clux Mate (r1c7/CluxMate)

Repo: [r1c7/CluxMate](https://github.com/r1c7/CluxMate) | Use cases: 4 | Page: https://dedupedai.com/use-cases/r1c7-cluxmate/

CluxMate is an open-source AI coding agent in Python that reads a codebase, plans changes, edits files, runs commands, and answers questions, through a headless CLI, a Textual TUI, or an Electron desktop app.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Run a one-shot coding fix from a script or CI job: https://dedupedai.com/use-cases/r1c7-cluxmate/one-shot-fix-in-ci-or-scripts/
- Run parallel agent tasks in separate git worktrees: https://dedupedai.com/use-cases/r1c7-cluxmate/run-parallel-tasks-in-separate-worktrees/
- Audit an agent's claims against its session log: https://dedupedai.com/use-cases/r1c7-cluxmate/audit-an-agent-session-from-its-log/
- Get a read-only code review from a reviewer subagent: https://dedupedai.com/use-cases/r1c7-cluxmate/get-a-read-only-review-from-a-reviewer-subagent/

## How to run a one-shot coding fix from a script or CI job with Clux Mate

Repo: [r1c7/CluxMate](https://github.com/r1c7/CluxMate) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Send one prompt to the headless CLI and review the file changes it makes.

**Who it is for:** A developer or CI maintainer who wants an agent to fix a small, well-defined problem in a repository without opening an interactive session.

**The problem:** Routine fixes such as a failing test or a small refactor take time to set up interactively, and scripts need a non-interactive way to call an agent.

**The outcome:** The agent runs one prompt, edits the files it decides to change, and reports a result that you can review as a diff.

### Steps

1. Install Clux Mate by following the install section of the README.
2. Point Clux Mate at your provider by setting its base_url to an OpenAI-compatible endpoint, as the README describes.
3. Run `cluxmate -p "Fix the failing unit tests in <path> and summarize what you changed"` from the repository root. (inferred, check the README)
4. Run `git diff` and read every changed file before you commit anything. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the CluxMate repo (r1c7/CluxMate) to run the headless command cluxmate -p with a prompt that targets <path>. Expected output: a short summary from the agent and a set of changed files. Verify with git diff that only files inside <path> changed and that the tests named in the prompt now pass.
```

### Caveat

The README excerpt does not show the config file format or install commands, so confirm those in the full README before the first run.

Risk, credentials: The API key for your provider must be supplied to Clux Mate, so store it in your CI secret store and not in the repository.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/r1c7-cluxmate/one-shot-fix-in-ci-or-scripts/

## How to run parallel agent tasks in separate git worktrees with Clux Mate

Repo: [r1c7/CluxMate](https://github.com/r1c7/CluxMate) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Give each task its own linked worktree and branch so sessions stop editing the same files.

**Who it is for:** A developer who has two or more tasks in the same repository and wants agent sessions to run at the same time without conflicts.

**The problem:** When several agent sessions edit the same working directory, their changes collide and are hard to separate.

**The outcome:** Each session works in its own linked git worktree on its own branch, and deleting a session can remove its tree.

### Steps

1. Open the project in the Textual TUI by running `cluxmate` from the repository root.
2. Create a session worktree from the desktop sidebar, or run `cluxmate worktree …` with the arguments for your branch.
3. Give each worktree its own task, since inside a tree the composer does not offer branch switches.
4. When a task ends, delete its session and choose whether the worktree goes with it.

### Prompt to paste into your agent

```text
Use the CluxMate repo (r1c7/CluxMate) to create a separate worktree for the task <task name> with cluxmate worktree. Expected output: a new linked git worktree on its own branch. Verify with git worktree list that the new tree exists and that the main working directory is unchanged.
```

### Caveat

Worktrees depend on git, so the project must already be a git repository, and the README excerpt does not list the full argument set for cluxmate worktree.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/r1c7-cluxmate/run-parallel-tasks-in-separate-worktrees/

## How to audit an agent's claims against its session log with Clux Mate

Repo: [r1c7/CluxMate](https://github.com/r1c7/CluxMate) | Difficulty: Advanced | Time: 1 hour | Basis: From the README

Check whether an agent's 'done' matches the tool calls that actually ran.

**Who it is for:** A reviewer or team lead who needs to confirm that an agent's summary of its work is backed by real tool calls and file changes.

**The problem:** An agent's final message describes what it believes happened, which may not match the commands and writes that actually ran.

**The outcome:** You can locate the completion audit outcome in the turn's log entry and confirm the claims against the filesystem.

### Steps

1. Run a task in the TUI with `cluxmate`, or in one shot with `cluxmate -p "<prompt>"`. (inferred, check the README)
2. Open the session's append-only `.jsonl` event log.
3. Find the `turn/end` event for the turn and read its `reason.completion_audit` field.
4. For each claim that files changed, check the filesystem and the `git diff` yourself. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the CluxMate repo (r1c7/CluxMate) to run the task <task> and then inspect its session .jsonl log. Expected output: the turn/end entry with completion_audit and a list of claims checked. Verify that every claimed file change appears in git diff.
```

### Caveat

The README excerpt describes the log events but not where session files are stored, so locate the session directory from the full README or the app before starting.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/r1c7-cluxmate/audit-an-agent-session-from-its-log/

## How to get a read-only code review from a reviewer subagent with Clux Mate

Repo: [r1c7/CluxMate](https://github.com/r1c7/CluxMate) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Delegate a review to a subagent that is restricted to evidence and cannot edit files.

**Who it is for:** A developer who wants a second look at a change before merging, without the reviewer modifying the code.

**The problem:** Asking the same agent that wrote a change to review it gives little independent check, and a reviewer that can edit may silently change the code.

**The outcome:** A reviewer report with a machine-readable status header, where unsupported success claims are downgraded to partial.

### Steps

1. Open the project in the TUI by running `cluxmate` from the repository root.
2. Ask the main agent to delegate the review to the `reviewer` subagent for the files you name.
3. Read the child report and its status header, which includes status, end reason, turns used, and elapsed time.
4. Check any issue the reviewer raises against the code yourself before acting on it.

### Prompt to paste into your agent

```text
Use the CluxMate repo (r1c7/CluxMate) to ask the reviewer subagent to review <files or diff>. Expected output: a review report with a status header. Verify that the reviewer made no file edits by checking git status before and after.
```

### Caveat

The README describes the reviewer as evidence-gated and fixing nothing, but it does not say how to invoke a subagent from the CLI, so the invocation runs through the interactive agent.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/r1c7-cluxmate/get-a-read-only-review-from-a-reviewer-subagent/