# Feedback

Report bugs, request features, or ask questions — directly from your development environment. The feedback tool collects environment context automatically and opens a pre-filled GitHub issue in your browser.

No tokens or API keys required. You authenticate with your own GitHub account when you submit.

## Quick start

From any AI coding agent with the sitemd MCP server:

```bash
/feedback bug the dev server crashes on empty pages
```

Or from the CLI:

```bash
sitemd feedback bug the dev server crashes on empty pages
```

Both open your browser to a GitHub issue form pre-filled with the title, a `bug` label, and your environment details (sitemd version, Node version, OS).

## How it works

The feedback tool builds a GitHub new-issue URL with query parameters for title, body, and labels. It appends an environment block to the issue body automatically:

```text
### Environment
- sitemd: v1.0.0
- Node: v22.0.0
- OS: darwin 24.0.0 (arm64)
```

When called from the CLI, it opens the URL in your default browser. When called via MCP, the agent receives the URL and presents it to you.

## Issue types

Pass the type as the first argument:

| Type | Label applied | When to use |
| --- | --- | --- |
| `bug` | `bug` | Something is broken or behaves unexpectedly |
| `feature` | `enhancement` | A new capability or improvement you want |
| `question` | none | General questions about how something works |

If you omit the type, it defaults to `question`.

## CLI command

```bash
sitemd feedback [type] [title]
```

Examples:

```bash
sitemd feedback                              # blank issue form with env context
sitemd feedback bug build fails with no pages
sitemd feedback feature support for dark mode toggle
sitemd feedback question how do groups work
```

The CLI prints your environment info to the terminal and opens the issue form in your browser.

## MCP tool

To report issues programmatically, [report issues on GitHub](https://github.com/sitemd-cc/sitemd/issues).

**Parameters:**

| Name | Required | Description |
| --- | --- | --- |
| `type` | yes | `bug`, `feature`, or `question` |
| `title` | yes | Short summary of the issue |
| `body` | no | Detailed description — for bugs, include what happened vs. what was expected |

**Returns:** Object with `url` (the pre-filled GitHub issue URL) and `context` (collected environment info).

The agent presents the URL for you to open. You can edit the pre-filled content on GitHub before submitting.

## The /feedback skill

The `/feedback` skill guides AI agents through the submission process:

1. Parses your arguments for type and title
2. Asks for details if not already clear from conversation context
3. For bugs encountered during the current session, auto-populates the body with error context
4. Opens a pre-filled GitHub issue and presents the URL

```bash
/feedback feature
/feedback bug the clone tool skips images on relative URLs
```

When an error occurred earlier in the conversation, the agent includes the error message and what was attempted — you don't need to re-explain it.

## Related

- [MCP Server](/docs/mcp-server) — all available MCP tools
- [CLI Config](/docs/cli-config) — the `sitemd` command and all CLI subcommands
- [Getting Started](/docs/getting-started) — project setup and first steps