---
description: Deep research on any topic with evidence-tier labeling
argument-hint: <topic>
allowed-tools: mcp__video-research__web_search, mcp__video-research__research_deep, mcp__video-research__research_plan, mcp__video-research__research_assess_evidence, Write, Glob, Read, Bash
model: sonnet
---

# Research: $ARGUMENTS

> For web-grounded deep research with the Gemini Deep Research Agent ($2-5/task, 10-20 min),
> use `/gr:research-deep` instead. This command uses offline analysis (free, instant).

Run a multi-phase deep research analysis with progressive memory saving and automatic evidence-network visualization.

## Phase 1: Research (run BOTH in parallel)

These two calls are independent — they hit different models with separate quotas. Issue both tool calls in a single turn:

1. `web_search(query="$ARGUMENTS")` — uses Gemini Flash with Google Search grounding
2. `research_deep(topic="$ARGUMENTS", scope="moderate", thinking_level="high")` — uses Gemini Pro

Do NOT wait for one to finish before starting the other.

## Phase 2: Present & Save Initial Results

1. Present findings organized by evidence tier:
   - **CONFIRMED** — Multiple independent sources agree
   - **STRONG INDICATOR** — Credible evidence with minor gaps
   - **INFERENCE** — Reasonable conclusion from indirect evidence
   - **SPECULATION** — Plausible but unverified
   - **UNKNOWN** — Insufficient evidence
2. Highlight open questions and methodology critique

3. **Immediately save initial results**:
   a. Determine the memory directory: find the `.claude/` project memory path for the current working directory. Use `Glob` on `~/.claude/projects/*/memory/` to find the active project memory path if needed.
   b. Generate a slug from the topic: lowercase, hyphens, no special chars, max 50 chars (e.g., "impact of mcp on ai agents" → `impact-of-mcp-on-ai-agents`)
   c. Use `Write` to save at `<memory-dir>/gr/research/<slug>/analysis.md`:

```markdown
---
source: web research
topic: "$ARGUMENTS"
analyzed: <ISO 8601 timestamp>
updated: <ISO 8601 timestamp>
scope: moderate
findings_count: <number>
evidence_tiers:
  confirmed: <count>
  strong_indicator: <count>
  inference: <count>
  speculation: <count>
  unknown: <count>
---

# $ARGUMENTS

> Researched on <YYYY-MM-DD HH:MM>
> Scope: moderate

## Executive Summary  <!-- <YYYY-MM-DD HH:MM> -->

<2-3 sentence summary>

## Findings  <!-- <YYYY-MM-DD HH:MM> -->

### CONFIRMED
1. **<Finding>** — <evidence summary>

### STRONG INDICATOR
2. **<Finding>** — <evidence summary>

### INFERENCE
3. **<Finding>** — <evidence summary>

### SPECULATION
4. **<Finding>** — <evidence summary>

### UNKNOWN
5. **<Finding>** — <evidence summary>

## Sources  <!-- <YYYY-MM-DD HH:MM> -->

<Cited sources with URLs where available>

## Open Questions  <!-- <YYYY-MM-DD HH:MM> -->

<Unresolved questions for future investigation>

## Methodology Critique  <!-- <YYYY-MM-DD HH:MM> -->

<Assessment of research methodology limitations>
```

   d. Tell the user: **Saved initial research to `gr/research/<slug>/`**

## Phase 3: Enrich with Evidence Network

1. Map the research findings into a network structure:
   - Each finding becomes a node with its evidence tier
   - Open questions become nodes (distinct style)
   - Supporting/contradicting relationships between findings become edges
   - Shared sources create implicit "related to" edges between findings

2. Append an Evidence Network section to `analysis.md`:

```markdown
## Evidence Network  <!-- <YYYY-MM-DD HH:MM> -->

### Nodes
- **Finding 1** (CONFIRMED) — <claim summary>
- **Finding 2** (INFERENCE) — <claim summary>
- **Open Question 1** — <question>

### Relationships
- Finding 1 → *supports* → Finding 3
- Finding 2 → *contradicts* → Finding 4
- Open Question 1 → *challenges* → Finding 2
```

3. Update the `updated` timestamp in frontmatter.

## Phase 4: Background Visualization (optional)

Ask the user with `AskUserQuestion`:
- Question: "Generate interactive evidence network visualization? (runs in background)"
- Option 1: "Yes (Recommended)" — description: "HTML visualization + screenshot + workspace copy, runs asynchronously"
- Option 2: "Skip" — description: "Finish now with research only"

**If yes**: Spawn the `visualizer` agent in the background with this prompt:
```
analysis_path: <memory-dir>/gr/research/<slug>/analysis.md
template_name: research-evidence-net
slug: <slug>
content_type: research
```

**Do NOT wait** for the visualizer. Continue immediately to Deeper Analysis below. The user will be notified when visualization is done.

**If skip**: Confirm: **Research complete — saved to `gr/research/<slug>/`**
- `analysis.md` — timestamped findings with evidence tiers

## Deeper Analysis

If the user wants deeper analysis, offer:
- Re-run with scope="deep" or "comprehensive" — results append to existing analysis.md with new timestamps
- Verify specific claims with `research_assess_evidence` — append verification results
- Broader context with `web_search` for specific sub-topics

Any deeper analysis appends timestamped sections to the existing `analysis.md` and may trigger a visualization update.