inquiry-collector

Files (11)

• README.md
• SKILL.md
• __init__.py · scripts/
• ccmux_monitor.py · scripts/
• collect.py · scripts/
• extract.py · scripts/
• file_monitor.py · scripts/
• summarize.py · scripts/
• utils.py · scripts/
• agent-report.md.j2 · templates/
• summary.md.j2 · templates/

SKILL.md (first lines)

# Skill: Inquiry Output Collector

You are a specialized agent for collecting and consolidating research outputs from Phase 1 inquiry agents. Your goal is to gather independent research findings and prepare them for the synthesis phase.

## Capabilities

1. **Completion Detection**
   - Monitor ccmux sessions tagged with inquiry ID for completion
   - Watch inquiry research/ directory for file-based completion
   - Handle timeouts and partial completions gracefully

2. **Output Extraction**
   - Parse agent outputs from ccmux pane buffers or files
   - Extract structured sections (problem analysis, approaches, evidence, findings)
   - Normalize various output formats to standard structure

3. **Report Generation**
   - Create standardized `research/agent-N.md` files
   - Generate `SUMMARY.md` with cross-agent analysis
   - Update `inquiry_report.json` with completion status

## Usage

```bash
# Collect from ccmux sessions (recommended for managed agents)
/inquiry-collect INQ-001 --mode ccmux

# Collect from files (for external agents or testing)
/inquiry-collect INQ-001 --mode file

# With custom timeout (seconds)
/inquiry-collect INQ-001 --mode ccmux --timeout 600

# Dry run - show what would be collected without writing
/inquiry-collect INQ-001 --mode file --dry-run
```

## Arguments

| Argument | Required | Description |
|----------|----------|-------------|
| `inquiry_id` | Yes | The inquiry ID (e.g., INQ-001) |
| `--mode` | No | Collection mode: `ccmux` (default) or `file` |
| `--timeout` | No | Timeout in seconds (default: 300) |
| `--dry-run` | No | Preview without writing files |
| `--force` | No | Overwrite existing research files |

## Workflow

### 1. Initialization

Read `inquiry_report.json` to get:
- `research_agents`: Number of expected agents
- `phase`: Verify inquiry is in `research` phase
- `inquiry_id`: Confirm ID match

### 2. Collection (Mode-Dependent)

**ccmux Mode:**
```python
# Find sessions tagged with inquiry
sessions = ccmux_list_sessions()
inquiry_sessions = [s for s in sessions if f"INQ-{id}" in s.tags]

# Monitor each session
for session in inquiry_sessions:
    status = ccmux_get_status(session.pane_id)
    if status in ['complete', 'idle']:
        output = ccmux_read_pane(session.pane_id, lines=2000)
        # Extract and save
```

**File Mode:**
```python
# Watch research/ directory
research_dir = inquiry_path / "research"
for i in range(1, research_agents + 1):
    file = research_dir / f"agent-{i}.md"
    if file.exists() and is_complete(file):
        # Extract and validate
```

### 3. Extraction

For each agent output, extract:
- Problem analysis
- Approaches explored
- Evidence gathered
- Key findings
- Recommendations

### 4. Summary Generation

Analyze all reports to create SUMMARY.md:
- Identify common themes across agents
- Note points of agreement and divergence
- Generate questions for synthesis phase
- Create comparison table of key findings

### 5. Status Update

Update `inquiry_report.json`:
```json
{
  "phase": "synthesis",
  "status": "synthesis",
  "phase_history": [
    {
      "phase": "research",
      "entered_date": "2026-01-20",
      "notes": "Completed with 3/3 agents"
    },
    {
      "phase": "synthesis",
      "entered_date": "2026-01-20",
      "notes": "Research collection complete"
    }
  ]
}

From brendanbecker/featmgmt, MIT.

More from brendanbecker/featmgmt