discussion-qa

# discussion-qa

Quality assurance specialist participant for discussions

## Installation

This is a CmdForge tool. Install it from the registry:

```bash
cmdforge install discussion-qa
```

## Usage

This tool is designed to be used as a discussion participant, typically called by an orchestration system. It accepts discussion context via stdin and returns a structured response with a comment and vote.

```bash
# Basic usage with discussion context
echo '{"topic": "...", "messages": [...]}' | discussion-qa

# With specific callout or @mention
echo '{"topic": "...", "messages": [...]}' | discussion-qa --callout "@qa What are the edge cases?"

# Custom paths for templates and diagrams
echo '{"topic": "...", "messages": [...]}' | discussion-qa --templates-dir ./custom/templates --diagrams-dir ./output/diagrams

# With progress logging
echo '{"topic": "...", "messages": [...]}' | discussion-qa --log-file qa-progress.log
```

## Arguments

| Flag | Default | Description |
|------|---------|-------------|
| `--callout` | _(empty)_ | Specific question or @mention context |
| `--templates-dir` | `templates` | Path to templates directory |
| `--diagrams-dir` | `diagrams` | Path to save diagrams |
| `--log-file` | _(empty)_ | Path to log file for progress updates |

## How It Works

The tool processes discussion context through an 8-step pipeline:

1. **Phase Detection** - Analyzes the discussion to determine the current phase and extract relevant context
2. **Diagram Generation** - Creates visual diagrams to aid in quality analysis (saved to diagrams directory)
3. **Progress Logging** - Logs initial progress if log file is specified
4. **AI Analysis** - Sends context to the `codex` provider for quality assurance analysis
5. **Progress Update** - Logs analysis completion
6. **JSON Extraction** - Parses the AI response to extract structured data
7. **Response Structuring** - Formats the comment and vote from the analysis
8. **Final Assembly** - Prepares the final JSON response with comment and vote

The tool outputs a JSON object containing the QA specialist's comment and vote on the discussion topic.

## Dependencies

None