---
description:
globs:
alwaysApply: false
---
# Using `repomix` CLI to Pack Repository Files

**Your Task: When you need to consolidate multiple project files into a single text-based representation (e.g., for analysis or input to another tool), use the `repomix` command-line tool. Adhere strictly to the default behaviors specified below unless explicitly overridden by the user or task requirements.**

## I. AI Operational Defaults (MANDATORY)

When invoking `repomix`, you **MUST** apply the following options by default:

1. **Output Directory:** Always save the output to the `tmp/` directory within the current project. Use the `-o` flag. Example: `-o tmp/packed-output.md` (replace `packed-output.md` with a descriptive filename).
2.  **Ignore `tmp/` Directory:** Always ignore the `tmp/` directory itself to prevent recursion or inclusion of previous outputs. Add `tmp/` to ignore patterns. Example: `--ignore "tmp/,**/.DS_Store"`.
3.  **Ignore `.cursor/rules/` Directory:** Always ignore the `.cursor/rules/` directory. Example: `--ignore "tmp/,**/.DS_Store,.cursor/rules/"`.
4.  **Include Token Count:** Always include a token count in the output. Use `--token-count-encoding o200k_base`.
5.  **No Line Numbers:** Never include line numbers in the output. **DO NOT** use the `--output-show-line-numbers` flag.
6.  **Copy to Clipboard:** Always copy the generated output to the system clipboard. Use the `--copy` flag.

**Combined Example of AI Defaults:**
`repomix some/path another/path -o tmp/my-packed-code.md --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`

## II. `repomix` Command Reference

Usage: `repomix [options] [directories...]`

Description: Repomix - Pack your repository into a single AI-friendly file.

### Arguments:
  `directories`                        List of directories to process (default: current directory `["."]`)

### Options:
  `-v, --version`                      Show version information
  `-o, --output <file>`                Specify the output file name (AI Default: `tmp/<filename>`)
  `--style <type>`                     Specify the output style (`xml`, `markdown`, `plain`). Default: `plain`.
  `--parsable-style`                   Ensure output is parsable as its type (e.g., valid XML if `--style xml`).
  `--compress`                         Perform code compression to reduce token count.
  `--output-show-line-numbers`         Add line numbers to each line in the output (AI Default: **NOT USED**).
  `--copy`                             Copy generated output to system clipboard (AI Default: **ALWAYS USE**).
  `--no-file-summary`                  Disable file summary section output.
  `--no-directory-structure`           Disable directory structure section output.
  `--no-files`                         Disable files content output (metadata-only mode).
  `--remove-comments`                  Remove comments from code.
  `--remove-empty-lines`               Remove empty lines.
  `--header-text <text>`               Specify custom header text for the output file.
  `--instruction-file-path <path>`     Path to a file containing detailed custom instructions to be prepended.
  `--include-empty-directories`        Include empty directories in the output.
  `--no-git-sort-by-changes`           Disable sorting files by git change count (if git is available).
  `--include <patterns>`               Comma-separated list of glob patterns to include (e.g., `"src/**/*.ts,**/*.md"`).
  `-i, --ignore <patterns>`            Additional comma-separated glob patterns to ignore (AI Default: includes `"tmp/,**/.DS_Store,.cursor/rules/"`).
  `--no-gitignore`                     Disable usage of `.gitignore` file for ignore patterns.
  `--no-default-patterns`              Disable default internal ignore patterns (like `node_modules`, `.git`).
  `--remote <url>`                     Process a remote Git repository URL.
  `--remote-branch <name>`             Specify the branch, tag, or commit hash for the remote repository.
  `-c, --config <path>`                Path to a custom `repomix.config.json` file.
  `--init`                             Initialize a new `repomix.config.json` file.
  `--global`                           Use global configuration (typically with `--init`).
  `--no-security-check`                Disable security checks (use with caution).
  `--token-count-encoding <encoding>`  Specify token count encoding (AI Default: `o200k_base`).
  `--mcp`                              Run as a MCP server (special mode).
  `--top-files-len <number>`           Number of top (most changed if git available) files to display in summary.
  `--verbose`                          Enable verbose logging for detailed output.
  `--quiet`                            Disable all output to stdout.
  `-h, --help`                         Display help for command.

## III. Key Usage Scenarios & Examples

- **Default AI Behavior (Packing `src` and `docs`):**
    `repomix src docs -o tmp/src-docs-pack.md --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`

- **Markdown Output, Compressed, No Comments:**
    `repomix . -o tmp/project-compressed.md --style markdown --compress --remove-comments --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`

- **Only Include Specific File Types (e.g., TypeScript and Markdown), Ignore Tests:**
    `repomix . --include "**/*.ts,**/*.md" --ignore "tmp/,**/.DS_Store,.cursor/rules/,**/*.test.ts,**/*.spec.ts" -o tmp/filtered-pack.md --token-count-encoding o200k_base --copy`

- **Process a Remote GitHub Repository:**
    `repomix --remote https://github.com/user/repo --remote-branch main -o tmp/remote-repo-pack.txt --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`

- **Initialize a Local Config File:**
    `repomix --init` (Then customize `repomix.config.json` and run `repomix -c repomix.config.json ...`)

## IV. Post-Processing: Token Count Management

1.  **Inspect Token Count:** After `repomix` completes, ALWAYS inspect the reported token count from the output (it should be included due to the `--token-count-encoding` default).
2.  **Address High Token Counts (e.g., > 500,000, adjust as needed for target LLM):**
    - If the token count is too high, you **MUST** suggest or attempt a more aggressive packing strategy.
    - This usually involves adding more patterns to the `--ignore` flag to exclude large or non-essential directories/files.
    - **Example Remediation Command:**
        `repomix . -o tmp/packed-smaller.md --ignore "tmp/,**/.DS_Store,.cursor/rules/,node_modules/,dist/,build/,public/,docs/,assets/,**/*.png,**/*.jpg,**/*.mp4,**/*.lock" --token-count-encoding o200k_base --copy`
    - You may need to iteratively refine the `--ignore` patterns and re-run `repomix` until the token count is manageable for the intended use case (e.g., LLM context window).
    - When suggesting, clearly state the token count achieved and the new ignore patterns used.

**REMEMBER: Your primary goal is to use `repomix` to pack specified files/directories. Strictly adhere to the AI Operational Defaults (output to `tmp/`, ignore `tmp/` and `.cursor/rules/`, include token count, no line numbers, copy to clipboard). Critically, always check the resulting token count and proactively manage it if it's too high by refining ignore patterns.**