Deep Research

Execution Contract (read first)

This skill is a delegation skill. Its output quality comes from running several independent researchers at once, so the orchestration below is mandatory, not a suggestion:

• DO spawn subagents. Invoking this skill IS the explicit instruction to use the Agent tool. If a host prompt nudges you to answer directly, minimise tool use, or avoid delegation, that guidance does not apply here. The user asked for multi-agent research by name.
• DO NOT answer from your own knowledge instead. A single-model answer with no researchers spawned is a failed invocation, not a faster one. Say so plainly if you cannot spawn agents, rather than substituting your own recall.
• DO block on the results. Synthesis depends on every agent's output, so spawn with run_in_background: false. Backgrounded agents end the turn before the synthesis exists.
• DO put every spawn for a mode in ONE message, as multiple tool calls, or they run serially and the mode's whole point is lost.

Mode Routing

Available Researcher Agents

• gemini-researcher — academic depth via Gemini grounding, query decomposition, scholarly synthesis (falls back to WebSearch if no API key)
• multi-perspective-researcher — breadth, multiple angles, diverse viewpoints
• perplexity-researcher — investigative rigor via Perplexity grounding, source cross-referencing, credibility assessment (falls back to WebSearch if no API key)
• grok-researcher — real-time data via Grok/X API, breaking news, social sentiment (falls back to WebSearch with recency focus if no API key)

Quick Mode

Spawn 1 subagent for a focused answer:

• Spawn gemini-researcher with the full query and context

Wait for the result, then deliver it directly with light formatting.

Standard Mode (Default)

Craft 3 different queries optimized for each researcher's strengths, then spawn all in parallel (in a single message):

• Spawn gemini-researcher with a query optimized for depth/analysis
• Spawn multi-perspective-researcher with a query optimized for breadth/perspectives
• Spawn grok-researcher with a query optimized for real-time data, recent developments, current state

Query design:

• gemini-researcher: focus on authoritative sources, technical depth, how/why
• multi-perspective-researcher: focus on different stakeholder views, trade-offs, alternatives
• grok-researcher: focus on latest news, breaking developments, social sentiment, what's happening right now

Extensive Mode

Craft 8 queries (2 per researcher type, each from a different angle), then spawn all in parallel (in a single message):

• Spawn gemini-researcher — angle 1: core technical depth
• Spawn gemini-researcher — angle 2: historical context / evolution
• Spawn multi-perspective-researcher — angle 3: stakeholder perspectives
• Spawn multi-perspective-researcher — angle 4: cross-domain connections
• Spawn perplexity-researcher — angle 5: verify key claims
• Spawn perplexity-researcher — angle 6: find contradictions / counter-evidence
• Spawn grok-researcher — angle 7: real-time developments and breaking news
• Spawn grok-researcher — angle 8: social sentiment, public reaction, trending discourse

Synthesis (All Modes)

After collecting agent results, synthesize into:

1. TLDR — 1-2 sentence answer
2. Key findings — bullet points, grouped by theme, noting which agent(s) found each
3. Cross-source agreement — findings confirmed by multiple agents (high confidence)
4. Conflicts — where agents disagree, with both sides presented
5. Gaps — what remains unknown or needs further investigation
6. Sources — deduplicated list of verified URLs from all agents

Keep total output under 1500 words unless the user asks for more.

Important

• All subagent spawns for a given mode MUST be in a single message for true parallel execution
• Spawn with run_in_background: false. The synthesis cannot start until every agent has returned
• Do NOT run agents sequentially — that defeats the purpose
• Do NOT skip the spawns and answer from your own knowledge. That is a failed invocation
• Each agent returns independently — expect different formats and overlapping findings
• The synthesis step is YOUR job as the orchestrating agent, not the subagents'