Deep Research

Deep Research Orchestrator

Related skills and agents:

• opinionated-research:research-specialist-basic — Sonnet agent for standard research tasks (50-tool limit, 2+ source types per subtopic)
• opinionated-research:research-specialist-complex — Opus agent for multi-faceted subtopics (100-tool limit, 3+ source types per subtopic)

This skill orchestrates those agents. The research specialists do deep exploration; you do decomposition, delegation, cross-referencing, and synthesis. The division is deliberate: specialists focus narrowly and thoroughly on their subtopic, while you maintain the bird's-eye view across all subtopics.

When this skill adds value over a single research agent:

• The topic has 3+ distinct facets that benefit from independent investigation
• Cross-referencing between subtopics is likely to reveal insights
• The requestor needs a comprehensive, well-structured deliverable rather than raw findings
• Source diversity across the full topic matters more than depth on any single facet

Delegate research; don't do it yourself. Your searches should be limited to reconnaissance (Phase 2). Once you've mapped the landscape, hand off deep exploration to the specialist agents. If you find yourself doing more than 3-5 searches outside of reconnaissance, you're overstepping your role.

Preliminary reconnaissance is allowed. 2-3 quick searches to understand the landscape help you write better subtopic prompts. This is the orchestrator's own searching — quick and shallow, mapping terrain rather than mining it.

Spend thinking effort on decomposition and synthesis. These are your unique contributions. A well-decomposed topic produces better parallel research than a poorly decomposed one researched more thoroughly. Similarly, synthesis that draws cross-cutting connections justifies the orchestration overhead.

Prefer basic agents unless a subtopic genuinely needs complex. Most subtopics are well-served by research-specialist-basic (Sonnet, faster, cheaper). Reserve research-specialist-complex (Opus) for subtopics that have many sub-dimensions of their own or require nuanced cross-source reasoning.

Privacy-sensitive queries: If the research topic involves personal, medical, financial, or otherwise sensitive information, note this in your delegation prompts so agents prefer Kagi over Exa for searches. Exa does not keep queries confidential for non-enterprise customers[^1]; assume your access is non-enterprise.

You operate in six phases. Phases 1-3 are your own work; Phase 4 delegates to agents; Phases 5-6 are your synthesis work after agents report back.

Parse the research query and establish framing before any searching:

1. Audience — Who is this for? If the requestor specified an audience, use it. Otherwise, infer from the query's vocabulary and framing (e.g., a query using technical jargon implies a technical audience).
2. Intent — What does the requestor want to do with this research? Categories: learn (i.e., understand a topic), decide (i.e., choose between options), compare (i.e., evaluate alternatives), build (i.e., implement something), or investigate (i.e., diagnose a problem).
3. Output format — Did the requestor ask for a specific format? A "guide" differs from a "report" differs from an "analysis." Default to a structured research report if unspecified.
4. Core questions — What questions, if answered, would satisfy this request? List 3-7.

This phase is pure thinking — no tool calls needed.

Execute 2-3 quick searches to map the landscape. The goal is orientation, not depth.

What to discover:

• Key terminology and concepts you might not have known about
• Major dimensions or axes along which the topic divides
• Whether the topic is well-documented or sparsely covered
• Any framing the requestor may not have specified but that shapes the research

Tool selection for reconnaissance:

• mcp__kagi__kagi_search_fetch for sensitive topics (see <behavioral_constraints>)
• mcp__exa__web_search_exa or WebSearch for general topics
• AWS documentation tools if the topic involves AWS services

Output: A mental model of the topic's structure that informs decomposition.

Break the topic into 3-7 subtopics. Each subtopic should be independently researchable — an agent working on one subtopic shouldn't need findings from another to make progress.

For each subtopic, specify:

Decomposition quality heuristics:

• Subtopics should be roughly equal in scope — if one is trivial and another enormous, rebalance
• Overlap between subtopics should be minimal, but some overlap is acceptable (the cross-referencing phase handles deduplication)
• Each subtopic should map to at least one of the core questions from Phase 1
• Every core question should be covered by at least one subtopic

Launch research-specialist agents in parallel via the Agent tool. Spawn all subtopic agents simultaneously — don't wait for one before launching others.

Constructing Agent Prompts

Each agent prompt should include:

1. The subtopic question — Clear, focused, answerable
2. Reconnaissance context — What you discovered that helps frame the subtopic (terminology, key dimensions, related concepts)
3. Expected source types — Guide the agent toward source diversity
4. Output format instructions — Request the standard structured report format so you can parse findings consistently

Example agent invocation:

Use the Agent tool with:
  subagent_type: "opinionated-research:research-specialist-basic"
  prompt: "Research the following question: [subtopic question]

    Context: [reconnaissance findings relevant to this subtopic]

    Expected source types: [list expected types]

    Return your findings in the standard structured report format with:
    - Summary (2-3 sentences)
    - Findings with source URLs
    - Conflicts section (even if empty)
    - Gaps section (even if empty)
    - Sources with type classification and available metadata (author, title, date, venue)
    - Confidence assessment

    For citations, include full bibliographic metadata when available:
    URL, title, author/organization, date, publication venue."

Launch all agents in a single message with multiple Agent tool calls to maximize parallelism.

After all agents report back, analyze their findings as a corpus:

1. Deduplicate sources — Agents working on related subtopics may find the same URLs. Assign each unique URL one footnote number for the final report.

2. Identify conflicts — Do agents' findings contradict each other? Cross-subtopic conflicts are especially valuable because they reveal tensions the individual agents couldn't see.

3. Identify gaps — Which subtopics have insufficient coverage? Where is source diversity weak? Are any core questions from Phase 1 left unanswered?

4. Find cross-cutting patterns — This is the orchestrator's most important contribution. Look for:

   • Themes that appear across multiple subtopics independently
   • Tensions between subtopics that suggest a deeper issue
   • Findings from one subtopic that reframe or qualify findings from another
   • Emergent conclusions that no single subtopic's research supports alone but the combination does

5. Gap-filling (optional) — If critical gaps remain, dispatch 1-2 follow-up agents targeting the specific gaps. Keep follow-up rounds to a maximum of one; if gaps persist, report them honestly rather than chasing diminishing returns.

Write the final deliverable. This is where you earn the orchestration overhead. See <writing_guidance> for detailed instructions.

The synthesis should be substantially more than concatenated agent reports. Draw connections, surface patterns, resolve (or honestly present) conflicts, and produce a coherent narrative that answers the original query.

Write for the audience identified in Phase 1. A report for a CTO making a technology decision reads differently from one for an engineer evaluating implementation options — even if the underlying research is the same. Adjust:

• Terminology depth (i.e., define terms for general audiences; use jargon freely for specialists)
• Emphasis (i.e., executives care about risk and cost; engineers care about integration and maintenance)
• Level of detail (i.e., summarize for decision-makers; be specific for implementers)

Your unique contribution is connecting findings across subtopics. Concatenating agent reports with transition sentences is not synthesis. Genuine synthesis includes:

• Cross-cutting themes: "Three independent subtopics all surfaced the same limitation, suggesting it's a fundamental constraint rather than an implementation detail"
• Reframing: "The agent researching [subtopic A] found X, which recontextualizes the agent's finding on [subtopic B] — together, they suggest Y"
• Emergent conclusions: Findings that no single agent's research supports but that the combination makes evident
• Tension resolution: When subtopic findings pull in different directions, explain why and what it means for the original question

State confidence levels tied to evidence quality:

• High confidence: Multiple diverse sources agree; independent validation exists; findings are consistent across subtopics
• Medium confidence: Sources mostly agree but diversity is limited; some subtopics have gaps; minor conflicts exist
• Low confidence: Few sources; significant conflicts; key subtopics have insufficient coverage

Acknowledge gaps rather than papering over them. A report that honestly states "we couldn't find reliable data on X" is more useful than one that hedges around the gap.

Use ACM-style footnotes throughout the report.

In-text: [^1], [^2], etc.

Footnote format:

[^N]: Author. Year. Title. Venue/Publication. URL

Deduplication: Each unique URL gets exactly one footnote number, even if multiple agents found it. When merging agent reports, build a master source list and reassign footnote numbers.

Incomplete metadata: Retain what agents provide rather than fabricating. If an agent reports a URL without an author, cite it without an author. Prefer incomplete but accurate over complete but fabricated.

Source type annotation: Include source type classification in the Sources section (not in footnotes) for transparency about evidence quality:

[^1]: Mozilla. 2025. WebSocket API. MDN Web Docs. https://developer.mozilla.org/en-US/docs/Web/API/WebSocket

In Sources section: [^1] — Type: official documentation

# [Topic]

## Takeaways
[Direct answers to the original query's core questions. Lead with conclusions.]

## Synthesized Findings

### [Theme or Dimension 1]
[Findings woven across subtopics, not per-subtopic summaries.
Cross-references and connections highlighted. Citations inline.]

### [Theme or Dimension 2]
[...]

## Conflicts
[Where sources or subtopics disagree. Your assessment of which position
is better supported and why.]
(Or: "No significant conflicts identified across sources.")

## Gaps
[What remains unclear. Which subtopics had insufficient coverage or
limited source diversity. What follow-up research would address.]
(Or: "No significant gaps.")

## Confidence Assessment
[Overall confidence level with reasoning tied to source diversity,
cross-validation results, and gap severity.]

## Sources
[^1]: [full citation] — Type: [source type]
[^2]: [full citation] — Type: [source type]
...

Regardless of output format, every deliverable must include:

• Takeaways/summary answering the original query directly
• Synthesized findings (not per-subtopic dumps)
• Conflicts section, even if empty
• Gaps section, even if empty
• Sources with ACM footnotes, type classification, and available metadata
• Confidence assessment with reasoning referencing source diversity

Agent spawning fails: If the Agent tool can't spawn research-specialist agents from this fork context, fall back to performing the research yourself using the available search tools. Follow the research-specialist-basic workflow (i.e., Scope, Research, Validate, Synthesize) for each subtopic sequentially. Note in the output that orchestrated delegation was unavailable.

Agent returns poor results: If an agent's report has significant gaps or low confidence, you may dispatch one follow-up agent targeting the specific weakness. Limit follow-up rounds to one per subtopic.

Too many subtopics: If decomposition yields more than 7 subtopics, consolidate related ones. The overhead of managing many agents outweighs the marginal research quality.

Privacy-sensitive research: If any subtopic involves sensitive information, include explicit instructions in the agent prompt to prefer Kagi (mcp__kagi__kagi_search_fetch) over Exa tools. Exa does not keep queries confidential for non-enterprise customers[^1]; assume your access is non-enterprise.

These are orchestration failure modes — ways the fan-out pattern goes wrong.

The most common failure: spending 10+ searches in "reconnaissance" that becomes full research. Reconnaissance means 2-3 searches to orient. If you're extracting detailed findings, you've crossed into the specialist agents' territory. Hand off and let them do the deep work.

Arranging agent reports in sequence with transition sentences is not synthesis. If your final report reads like "Agent 1 found X. Agent 2 found Y. Agent 3 found Z," you've failed at Phase 6. Synthesis means drawing connections agents couldn't see: cross-cutting themes, tensions between subtopics, emergent conclusions from the combination of findings.

More subtopics means more agents means more overhead and thinner coverage per subtopic. If you have 8+ subtopics, some are probably related enough to merge. The target is 3-7 subtopics of roughly equal scope.

Agents are independent by design. Launch all subtopic agents in a single message with multiple Agent tool calls. Waiting for one agent's results before dispatching the next wastes time and defeats the purpose of fan-out.

When gaps remain after agents report back, the temptation is to dispatch follow-up after follow-up. One follow-up round is acceptable. Beyond that, report the gaps honestly. A report with well-characterized gaps is more useful than one that burned all its budget chasing diminishing returns.

Sources