Cursor rule

1---
2description:
3globs:
4alwaysApply: false
5---
6# Using `repomix` CLI to Pack Repository Files
7 
8**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.**
9 
10## I. AI Operational Defaults (MANDATORY)
11 
12When invoking `repomix`, you **MUST** apply the following options by default:
13 
141. **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).
152.  **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"`.
163.  **Ignore `.cursor/rules/` Directory:** Always ignore the `.cursor/rules/` directory. Example: `--ignore "tmp/,**/.DS_Store,.cursor/rules/"`.
174.  **Include Token Count:** Always include a token count in the output. Use `--token-count-encoding o200k_base`.
185.  **No Line Numbers:** Never include line numbers in the output. **DO NOT** use the `--output-show-line-numbers` flag.
196.  **Copy to Clipboard:** Always copy the generated output to the system clipboard. Use the `--copy` flag.
20 
21**Combined Example of AI Defaults:**
22`repomix some/path another/path -o tmp/my-packed-code.md --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`
23 
24## II. `repomix` Command Reference
25 
26Usage: `repomix [options] [directories...]`
27 
28Description: Repomix - Pack your repository into a single AI-friendly file.
29 
30### Arguments:
31  `directories`                        List of directories to process (default: current directory `["."]`)
32 
33### Options:
34  `-v, --version`                      Show version information
35  `-o, --output <file>`                Specify the output file name (AI Default: `tmp/<filename>`)
36  `--style <type>`                     Specify the output style (`xml`, `markdown`, `plain`). Default: `plain`.
37  `--parsable-style`                   Ensure output is parsable as its type (e.g., valid XML if `--style xml`).
38  `--compress`                         Perform code compression to reduce token count.
39  `--output-show-line-numbers`         Add line numbers to each line in the output (AI Default: **NOT USED**).
40  `--copy`                             Copy generated output to system clipboard (AI Default: **ALWAYS USE**).
41  `--no-file-summary`                  Disable file summary section output.
42  `--no-directory-structure`           Disable directory structure section output.
43  `--no-files`                         Disable files content output (metadata-only mode).
44  `--remove-comments`                  Remove comments from code.
45  `--remove-empty-lines`               Remove empty lines.
46  `--header-text <text>`               Specify custom header text for the output file.
47  `--instruction-file-path <path>`     Path to a file containing detailed custom instructions to be prepended.
48  `--include-empty-directories`        Include empty directories in the output.
49  `--no-git-sort-by-changes`           Disable sorting files by git change count (if git is available).
50  `--include <patterns>`               Comma-separated list of glob patterns to include (e.g., `"src/**/*.ts,**/*.md"`).
51  `-i, --ignore <patterns>`            Additional comma-separated glob patterns to ignore (AI Default: includes `"tmp/,**/.DS_Store,.cursor/rules/"`).
52  `--no-gitignore`                     Disable usage of `.gitignore` file for ignore patterns.
53  `--no-default-patterns`              Disable default internal ignore patterns (like `node_modules`, `.git`).
54  `--remote <url>`                     Process a remote Git repository URL.
55  `--remote-branch <name>`             Specify the branch, tag, or commit hash for the remote repository.
56  `-c, --config <path>`                Path to a custom `repomix.config.json` file.
57  `--init`                             Initialize a new `repomix.config.json` file.
58  `--global`                           Use global configuration (typically with `--init`).
59  `--no-security-check`                Disable security checks (use with caution).
60  `--token-count-encoding <encoding>`  Specify token count encoding (AI Default: `o200k_base`).
61  `--mcp`                              Run as a MCP server (special mode).
62  `--top-files-len <number>`           Number of top (most changed if git available) files to display in summary.
63  `--verbose`                          Enable verbose logging for detailed output.
64  `--quiet`                            Disable all output to stdout.
65  `-h, --help`                         Display help for command.
66 
67## III. Key Usage Scenarios & Examples
68 
69- **Default AI Behavior (Packing `src` and `docs`):**
70    `repomix src docs -o tmp/src-docs-pack.md --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`
71 
72- **Markdown Output, Compressed, No Comments:**
73    `repomix . -o tmp/project-compressed.md --style markdown --compress --remove-comments --ignore "tmp/,**/.DS_Store,.cursor/rules/" --token-count-encoding o200k_base --copy`
74 
75- **Only Include Specific File Types (e.g., TypeScript and Markdown), Ignore Tests:**
76    `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`
77 
78- **Process a Remote GitHub Repository:**
79    `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`
80 
81- **Initialize a Local Config File:**
82    `repomix --init` (Then customize `repomix.config.json` and run `repomix -c repomix.config.json ...`)
83 
84## IV. Post-Processing: Token Count Management
85 
861.  **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).
872.  **Address High Token Counts (e.g., > 500,000, adjust as needed for target LLM):**
88    - If the token count is too high, you **MUST** suggest or attempt a more aggressive packing strategy.
89    - This usually involves adding more patterns to the `--ignore` flag to exclude large or non-essential directories/files.
90    - **Example Remediation Command:**
91        `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`
92    - 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).
93    - When suggesting, clearly state the token count achieved and the new ignore patterns used.
94 
95**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.**
96