| name | internet-research-orchestrator |
| description | Orchestrate comprehensive TODAS research for novel/emerging domains (1-7 subagents adaptive). Specializes in unprecedented topics, post-training data, and emerging technologies. Uses adaptive depth-based methodology: straightforward queries (1 agent), standard queries (2-3 agents), complex queries (5-7 agents). Handles depth-first (multiple perspectives), breadth-first (distinct sub-topics), and straightforward investigations. Triggers include "novel", "emerging", "2025", "2026", "unprecedented", "new technology", research on topics that didn't exist during training cutoff. |
Novel Domain Research Orchestration (Tier 5 TODAS)
What is TODAS: Tactical Optimization & Depth-Adaptive System - an adaptive research methodology that adjusts agent count (1-7) and research depth based on query complexity and novelty. Optimized for emerging domains and post-training information.
Quick Start
- Analyze query novelty: Assess if topic is novel/emerging (post-training, unprecedented, 2025+ developments)
- Determine query type: Depth-first (multiple perspectives), Breadth-first (distinct sub-topics), or Straightforward
- Setup progress tracking: Use TodoWrite to create task list for research phases
- Calculate adaptive agent count: 1 (simple) to 7 (complex) based on query dimensions and novelty
- Execute TODAS workflow: Assessment → Query Type → Plan → Execution
- Spawn research-subagent instances: Use Task tool with adaptive count (1-7 agents in parallel)
- Synthesize findings: Integrate results from all subagents into coherent analysis
- Report completion: Summary with source attribution and novelty assessment
TODAS Methodology Overview
Tier 5 specialization: Novel and emerging domains requiring adaptive orchestration.
When to use this skill:
- ✅ Novel/emerging domains (didn't exist during training)
- ✅ Post-training developments (2025, 2026 technologies)
- ✅ Unprecedented topics (new paradigms, bleeding-edge tech)
- ✅ Rapidly evolving fields (AI agents, quantum computing, Web3)
- ✅ Multi-faceted queries requiring coordination (1-7 dimensions)
When NOT to use this skill:
- ❌ Established domains with known patterns (use Tier 4 internet-deep-orchestrator)
- ❌ Simple lookups (use Tier 1 web-researcher)
- ❌ Specialist queries (use Tier 2: academic-researcher, trend-analyst, etc.)
- ❌ Standard multi-dimensional research (use Tier 3 internet-light-orchestrator)
Adaptive depth: TODAS adjusts research depth dynamically:
- Straightforward queries: 1 subagent (direct investigation)
- Standard complexity: 2-3 subagents (multiple perspectives or sub-topics)
- Medium complexity: 3-5 subagents (multi-faceted approaches)
- High complexity: 5-7 subagents (broad coverage, many dimensions)
Tactical optimization: Stop research when diminishing returns reached (efficiency over completeness).
Workflow: 4-Phase TODAS Process
Phase 1: Assessment and Breakdown
Analyze the user's query thoroughly:
- Identify core concepts: Main entities, relationships, key questions
- List required data: Specific facts, temporal constraints, contextual boundaries
- Assess novelty: Is this topic post-training? Emerging? Unprecedented?
- User expectations: What form should answer take? (Report, analysis, comparison, list, etc.)
- Critical analysis: What features are most important? What does user care about most?
Output: Clear understanding of query scope, novelty level, and expected deliverable format.
Phase 2: Query Type Determination
Explicitly classify query into one of three types:
Depth-First Query
When: Multiple perspectives on SAME issue (going deep from many angles)
Characteristics:
- Single core question benefiting from diverse approaches
- Requires different viewpoints, methodologies, or sources
- Example: "What are the most effective treatments for depression?" (explore different treatments/approaches)
- Example: "What caused the 2008 financial crisis?" (economic, regulatory, behavioral, historical perspectives)
- Example: "Best approach to building AI finance agents in 2025?" (multiple methodologies)
Research strategy:
- Deploy 3-5 subagents exploring different methodological approaches
- Each subagent investigates from unique perspective
- Synthesis integrates diverse viewpoints into coherent analysis
Breadth-First Query
When: Distinct, independent sub-questions (going wide across topics)
Characteristics:
- Naturally divides into multiple parallel research streams
- Sub-topics can be researched independently
- Example: "Compare economic systems of three Nordic countries" (3 independent country researches)
- Example: "Fortune 500 CEOs net worths and names" (intractable as single thread, split into batches)
- Example: "Compare major frontend frameworks" (identify frameworks, then research each)
Research strategy:
- Enumerate all distinct sub-questions/sub-tasks
- Deploy subagents with clear, crisp boundaries (prevent overlap)
- Prioritize by importance and complexity
- Aggregate findings into coherent whole
Straightforward Query
When: Focused, well-defined, single investigation sufficient
Characteristics:
- Simple fact-finding or basic analysis
- Does not benefit from extensive multi-agent research
- Example: "What is current population of Tokyo?" (simple lookup)
- Example: "List all Fortune 500 companies" (single resource fetch)
- Example: "Tell me about bananas" (basic query, short answer expected)
Research strategy:
- Deploy 1 subagent with clear, focused instructions
- Specify exact data points required
- Include basic verification methods
- Synthesize findings efficiently
Output: Explicit query type classification with reasoning.
Phase 3: Detailed Research Plan Development
Based on query type, develop specific plan:
For Depth-First Queries:
- Define 3-5 different methodological approaches or perspectives
- List specific expert viewpoints or sources of evidence
- Plan how each perspective contributes unique insights
- Specify synthesis strategy for integrating findings
- Example: "What causes obesity?" → genetic factors, environmental influences, psychological aspects, socioeconomic patterns, biomedical evidence
For Breadth-First Queries:
- Enumerate all distinct sub-questions/sub-tasks
- Identify most critical sub-questions (focus on essential, avoid every angle)
- Prioritize by importance and expected complexity
- Define clear boundaries between sub-topics (prevent overlap)
- Plan aggregation strategy
- Example: "Compare EU country tax systems" → retrieve EU countries list, define comparison metrics, batch research by region (Northern, Western, Eastern, Southern Europe)
For Straightforward Queries:
- Identify most direct, efficient path to answer
- Determine if basic fact-finding or minor analysis needed
- Specify exact data points required
- Determine most relevant sources
- Plan basic verification methods
- Create extremely clear task description for subagent
For all query types, evaluate each step:
- Can this be broken into independent subtasks? (efficiency)
- Would multiple perspectives benefit this? (depth)
- What specific output is expected? (clarity)
- Is this strictly necessary to answer query? (focus)
Output: Concrete research plan with clear subagent allocation.
Phase 3a: Dimension Complexity Assessment (Resource Planning)
Before selecting specialists, assess each dimension's complexity to guide resource allocation:
Complexity Scoring Framework
For EACH dimension, evaluate these factors:
-
Sub-domains Count: How many distinct research areas?
- Example: Security dimension = academic papers + current threats + compliance standards = 3 sub-domains
- Scoring: 1 sub-domain = +1, 2 sub-domains = +2, 3+ sub-domains = +3
-
Criticality Level: High-stakes domain requiring extra rigor?
- HIGH criticality: Security, medical, financial, legal, compliance
- MODERATE criticality: Business strategy, market analysis
- LOW criticality: General information, non-critical topics
- Scoring: HIGH = +2, MODERATE = +1, LOW = +0
-
Novelty/Uncertainty: Emerging or post-training topic?
- HIGH novelty: 2025+ developments, unprecedented topics, emerging tech
- MODERATE novelty: Recent 2024 developments, evolving fields
- LOW novelty: Established topics, well-documented areas
- Scoring: HIGH = +1, MODERATE = +0.5, LOW = +0
-
Source Diversity Needed: How many different source types required?
- Examples: Academic journals, industry blogs, official docs, market reports, standards/regulations
- Scoring: 1-2 source types = +2, 3+ source types = +3
Complexity Classification (total score):
- Simple (1-3 points): 1 specialist sufficient
- Moderate (4-6 points): 1-2 specialists recommended
- Complex (7-9 points): 2-3 specialists recommended
- Critical (10+ points): 3+ specialists + dedicated fact-checker REQUIRED
Example Complexity Assessment
Dimension 4: Security and Multi-Tenancy Architectures
Factors:
- Sub-domains: Academic security papers + Current 2025 threats + Compliance standards = 3 sub-domains → +3
- Criticality: HIGH (security domain) → +2
- Novelty: MODERATE (2025 emerging security patterns) → +0.5
- Source diversity: Academic + Industry + Regulatory = 3 types → +3
Total Score: 3 + 2 + 0.5 + 3 = 8.5 points → COMPLEX
Recommendation: 2-3 specialists for comprehensive coverage
- Consider: academic-researcher (papers) + web-researcher (threats) + search-specialist (standards)
Dimension 1: Mobile-Native Push Notifications
Factors:
- Sub-domains: Platform implementations only = 1 sub-domain → +1
- Criticality: LOW (informational) → +0
- Novelty: LOW (established patterns) → +0
- Source diversity: Official docs + Engineering blogs = 2 types → +2
Total Score: 1 + 0 + 0 + 2 = 3 points → SIMPLE
Recommendation: 1 specialist sufficient
- Likely match: web-researcher (current platform documentation)
Output: Complexity score and allocation recommendation for each dimension.
Phase 3b: Specialist Selection with Self-Challenge (CRITICAL QUALITY GATE)
For EACH dimension, perform rigorous specialist selection with adversarial validation:
Step 1: Requirements Analysis & Initial Selection
Dimension [N]: [Dimension Name]
Requirements Analysis:
- What information is needed? (facts, trends, papers, market data, standards)
- What sources are required? (academic journals, industry blogs, official docs, regulations)
- Current vs future focus? (2024 current state vs 2025+ emerging trends)
- Theoretical vs production focus? (research papers vs real-world implementations)
Candidate Specialists (list 2-3 types that COULD handle this dimension):
- Candidate A: [specialist_type] - Strengths: [what they excel at for this dimension]
- Candidate B: [specialist_type] - Strengths: [what they excel at for this dimension]
- Candidate C: [specialist_type] - Strengths: [what they excel at for this dimension]
Initial Selection: [specialist_type]
Initial Rationale: [Why this specialist best matches the dimension requirements - be specific about the match between requirements and specialist capabilities]
Step 2: Self-Challenge Phase (🚨 MANDATORY - DO NOT SKIP)
The self-challenge phase prevents lazy defaulting and ensures optimal specialist matching.
🚨 CHALLENGE: "Wait, why didn't I choose [alternative_specialist] instead of [initial_selection]?"
For EACH alternative candidate (repeat 2-3 times per dimension):
Alternative: [alternative_specialist_type]
Gains if chosen:
- What unique value would this specialist provide?
- What perspectives/sources/capabilities does it have that initial selection lacks?
- What dimension requirements would it serve BETTER?
- Example: "trend-analyst would provide emerging 2025 patterns and weak signal detection that web-researcher doesn't offer"
Cons if chosen:
- What would this specialist LACK compared to initial selection?
- What dimension requirements would be UNDERSERVED?
- What trade-offs would we accept?
- Example: "trend-analyst forecasts FUTURE trends but dimension needs CURRENT production implementations"
Comparison: [initial_selection] vs [alternative]
- Where initial wins: [Specific requirements where initial is stronger]
- Where alternative wins: [Specific requirements where alternative is stronger]
- Net assessment: [Which better matches the dimension's PRIMARY requirements?]
- 🚨 Critical question: Does this comparison reveal initial selection was suboptimal?
Verdict:
- ✅ KEEP [initial_selection] - Rationale: [Why initial still best after challenge]
- OR
- ❌ SWITCH to [alternative] - Rationale: [Why alternative is actually better - challenge caught a mismatch]
Repeat challenge for Alternative B, Alternative C
Step 3: Final Selection Documentation
Dimension [N] - FINAL SELECTION: [specialist_type]
Final Rationale (after surviving self-challenge):
- Chosen because: [Strengths that best match dimension requirements]
- Alternatives considered and rejected:
- [Alternative A]: Rejected because [specific weakness or mismatch for THIS dimension]
- [Alternative B]: Rejected because [specific weakness or mismatch for THIS dimension]
- Decision confidence: HIGH (explicit adversarial challenge performed and passed)
Example Self-Challenge Workflow
Dimension 5: Real-Time Coordination Mechanisms
Requirements:
- WebSocket/SSE/WebRTC protocol documentation
- Shopify Mobile Bridge architecture (current)
- W3C MiniApp standardization status
- Novel aspect: Query mentions "emerging in 2025" (cutting-edge focus)
Candidates:
- web-researcher: Current documentation, engineering blogs, official specs
- trend-analyst: Emerging real-time trends, future forecasts, 2025+ developments
- search-specialist: Deep protocol search, technical specifications
Initial Selection: web-researcher
Initial Rationale: Can access Shopify engineering blog, W3C specification documents, WebSocket/SSE/WebRTC official protocol documentation
🚨 CHALLENGE: "Wait, why didn't I choose trend-analyst instead?"
Alternative: trend-analyst
Gains:
- Query EXPLICITLY mentions "emerging in 2025" and "cutting-edge"
- trend-analyst excels at identifying future real-time coordination trends
- Could forecast WebTransport, WebRTC 2.0, bleeding-edge 2025+ protocols
- Weak signal detection for emerging patterns
Cons:
- Might miss current production implementations (Shopify Mobile Bridge is CURRENT, not future)
- Could over-focus on speculative technologies not yet production-ready
- W3C current standardization status needs CURRENT docs, not future speculation
Comparison: web-researcher vs trend-analyst
- Where web-researcher wins: Current production (Shopify), W3C current status (documentation)
- Where trend-analyst wins: Emerging 2025 protocols, future forecasting ("cutting-edge" keyword match)
- Net assessment: This dimension has BOTH current (Shopify, W3C status) AND future ("emerging 2025") aspects
- 🚨 WAIT: The "emerging in 2025" and "cutting-edge" keywords suggest future focus is PRIMARY!
Verdict: ❌ SWITCH to trend-analyst
Justification: Self-challenge revealed the "emerging 2025" and "cutting-edge" keywords indicate this is a FUTURE-focused dimension. trend-analyst's forecasting strength better matches the PRIMARY requirement (emerging patterns) than web-researcher's current documentation strength. Initial selection was suboptimal - self-challenge caught this mismatch.
Final Selection: trend-analyst ✅ (REVISED from web-researcher)
Final Rationale:
- Chosen because: "Emerging 2025" focus requires forecasting capability > current documentation
- web-researcher rejected: Strength is current state, but dimension emphasizes emerging/cutting-edge
- search-specialist rejected: Deep search less valuable than trend forecasting for future-focused dimension
- Decision confidence: HIGH (self-challenge caught keyword mismatch and corrected suboptimal initial choice)
Output: Final specialist selection per dimension with explicit challenge-survived rationales.
Phase 3c: Resource Allocation Challenge (Quantity per Dimension)
After selecting specialist TYPES, determine COUNTS (how many specialists per dimension):
Resource Allocation Framework
Default: 1 specialist per dimension (efficiency)
Upgrade to 2-3 specialists when:
- Complexity score ≥7 (Complex or Critical dimensions)
- Coverage analysis reveals significant gaps with single specialist
- Critical domain requires redundancy (security, medical, financial)
For EACH Dimension: Allocation Decision
Dimension [N]: [Dimension Name]
Complexity Score: [score from Phase 3a]
Selected Specialist(s): [type(s)]
Current Allocation: 1 specialist (default)
🚨 CHALLENGE: "Should this dimension get MORE than 1 specialist?"
Coverage Analysis (with current 1-specialist allocation):
- Specialist covers: [What research areas/sources this specialist will handle]
- Missing coverage: [What important areas/sources remain uncovered]
- Estimated coverage: [Percentage estimate, e.g., "40%" or "85%"]
- Risk assessment: [LOW/MODERATE/HIGH - Is the coverage gap acceptable?]
Allocation Options:
Option A: Keep 1 specialist (default)
- Coverage: [percentage]
- Cost: 1 agent
- Risk: [risk level] - [Explanation of what might be missed]
- Justification: [When is single specialist sufficient?]
Option B: Add 1 more specialist (upgrade to 2)
- Additional specialist: [type] covering [specific gaps]
- Coverage improvement: [X% → Y%]
- Cost: +1 agent (total research count: [N])
- Risk: [Reduced risk] - [How second specialist reduces gap]
- Justification: [Why second specialist worth the cost]
Option C: Add 2 more specialists (upgrade to 3)
- Additional specialists:
- [type1] covering [gaps]
- [type2] covering [gaps]
- Coverage improvement: [X% → 95%+]
- Cost: +2 agents (total research count: [N])
- Risk: ✅ LOW (comprehensive multi-source coverage)
- Justification: [Why CRITICAL dimension needs 3 specialists]
Decision: [Option A / B / C]
Allocation Justification: [Explicit reasoning for chosen option]
Example Resource Allocation Decision
Dimension 4: Security and Multi-Tenancy Architectures
Complexity Score: 8.5 (COMPLEX)
Selected Specialist: academic-researcher
Current Allocation: 1
🚨 CHALLENGE: "Should this dimension get MORE than 1 specialist?"
Coverage Analysis:
- academic-researcher covers: Security research papers, theoretical frameworks, academic studies
- Missing: Current 2025 threat landscape (industry reports, CVEs), Compliance standards (ISO 27001, SOC 2, GDPR)
- Estimated coverage: 40% (academic only, missing 60% of security dimension)
- Risk: 🔴 HIGH - Security is CRITICAL domain, 40% coverage unacceptable
Option A: Keep 1 specialist
- Coverage: 40%
- Cost: 1 agent
- Risk: 🔴 HIGH - Major gaps in threat landscape and compliance
- Justification: NOT ACCEPTABLE for security domain
Option B: Add web-researcher (2 specialists total)
- Additional: web-researcher covering current 2025 threats, CVE databases, industry security reports
- Coverage: 40% → 70%
- Cost: +1 agent (total 6 for research)
- Risk: ⚠️ MODERATE - Still missing compliance/standards coverage
- Justification: Improvement but still gaps
Option C: Add web-researcher + search-specialist (3 specialists total)
- Additional:
- web-researcher: Current threats, CVEs, industry reports (40% → 70%)
- search-specialist: ISO 27001, SOC 2, GDPR standards, compliance docs (70% → 95%)
- Coverage: 40% → 95% (comprehensive)
- Cost: +2 agents (total 7 for research)
- Risk: ✅ LOW - Comprehensive coverage across papers, threats, compliance
- Justification: Security is CRITICAL. 95% coverage justifies +2 agents. Meets professional security audit standards.
Decision: Option C (3 specialists for this dimension)
Allocation Justification: Security domain criticality + 40% baseline coverage gap = REQUIRE comprehensive 3-specialist approach. Cost (+2 agents) justified by risk reduction (HIGH → LOW) and professional coverage standard (95%).
Dimension 1: Mobile-Native Push Notifications
Complexity Score: 3 (SIMPLE)
Selected Specialist: web-researcher
Current Allocation: 1
🚨 CHALLENGE: "Should this dimension get MORE than 1 specialist?"
Coverage Analysis:
- web-researcher covers: Platform docs (iOS/Android), engineering blogs (Shopify, Gojek), official documentation (Apple, Google)
- Missing: Limited academic theory (but not needed for production-focused dimension)
- Estimated coverage: 85%
- Risk: ✅ LOW - Simple dimension, single specialist provides strong coverage
Option A: Keep 1 specialist
- Coverage: 85%
- Cost: 1 agent
- Risk: ✅ LOW - Minor academic gap not relevant to production focus
- Justification: Simple dimension, 85% coverage sufficient
Decision: Option A (1 specialist)
Allocation Justification: Simple dimension (score 3) with production focus. web-researcher provides 85% coverage. Academic gap irrelevant. Single specialist efficient and sufficient.
Output: Allocation plan with specialist count per dimension (most stay at 1, critical dimensions upgrade to 2-3).
Phase 3d: Budget Optimization & Final Verification
After all allocation decisions complete, perform final budget validation and repetition challenge:
Budget Summary
Total Research Allocation:
- Dimension 1: [X] specialist(s) - [types]
- Dimension 2: [X] specialist(s) - [types]
- Dimension 3: [X] specialist(s) - [types]
- Dimension 4: [X] specialist(s) - [types]
- Dimension 5: [X] specialist(s) - [types]
- ...
- Total Specialists: [count]
- Fact-Checkers: [count] (1 per critical dimension)
- Grand Total: [count] agents
Budget Status
Target Range: 5-7 agents (optimal for novel domain depth)
Maximum: 10 agents (cost ceiling)
Status:
- ✅ Within Target (5-7): Optimal allocation for novel domain research
- ⚠️ Above Target (8-10): Justification required - explain why complexity necessitates additional agents
- 🔴 Overrun (>10): MUST optimize or provide exceptional justification
Repetition Challenge (Quality Gate for Lazy Defaults)
🚨 CHALLENGE: "Did I default to [specialist] out of laziness rather than intentional strategy?"
If ANY specialist type used MORE than 1 time, re-examine EACH usage:
Specialist Type: [type used multiple times, e.g., "web-researcher"]
Used For: Dimension [A], Dimension [B], Dimension [C]
For EACH Dimension:
Dimension [A]: [Dimension Name]
- Why this specialist?: [Original rationale from Phase 3b]
- 🚨 Repetition Challenge: "Did I choose web-researcher because it's GENUINELY optimal, or because I already chose it for Dimension B and defaulted to familiarity?"
- Fresh Comparison:
- web-researcher provides: [specific value for THIS dimension]
- Alternative ([other specialist]) would provide: [what alternative offers]
- Net assessment: Which BETTER matches Dimension A's PRIMARY requirements?
- Verdict:
- ✅ REPETITION JUSTIFIED: [Why web-researcher genuinely optimal for THIS specific dimension, independent of other dimensions]
- ❌ LAZY DEFAULT DETECTED: Switch to [alternative] - [Why alternative actually better match]
Example Repetition Challenge:
Specialist Type: web-researcher
Used For: Dimension 1 (Push Notifications), Dimension 3 (Infrastructure), Dimension 5 (Coordination)
Dimension 1: Push Notifications
- Why web-researcher: Platform docs (iOS/Android), engineering blogs (Shopify, Gojek)
- 🚨 Challenge: "Genuine or lazy default?"
- Fresh comparison: web-researcher (platform docs) vs academic-researcher (push theory)
- Net: Platform docs > theory for production focus
- Verdict: ✅ JUSTIFIED - Production implementation focus requires platform docs
Dimension 3: Infrastructure
- Why web-researcher: Current cloud providers (AWS, GCP), infrastructure blogs
- 🚨 Challenge: "Genuine or lazy default?"
- Fresh comparison: web-researcher (current providers) vs trend-analyst (emerging infrastructure patterns)
- Net: Query asks "current state", not "2025 trends"
- Verdict: ✅ JUSTIFIED - "Current state" keyword = web-researcher correct
Dimension 5: Real-Time Coordination
- Why web-researcher: WebSocket docs, real-time protocols
- 🚨 Challenge: "Genuine or lazy default?"
- Fresh comparison: web-researcher (protocols) vs academic-researcher (coordination algorithms, research papers)
- Net: Query dimension includes "emerging 2025 approaches" → academic papers valuable
- Verdict: ❌ LAZY DEFAULT - Switch to academic-researcher (coordination research papers better match emerging focus)
Outcome: Dimension 5 revised from web-researcher → academic-researcher (repetition challenge caught lazy default)
Final Diversity Assessment (Not Enforced, But Informative)
Unique Specialist Types: [count]
Total Specialists: [count]
Diversity Ratio: [unique / total]
Assessment:
- ✅ High Diversity (≥80%): Strong variety, low repetition
- ⚠️ Moderate Diversity (50-79%): Some repetition, verify justified
- 🟡 Low Diversity (<50%): Significant repetition, ensure NOT lazy defaults
Note: Low diversity acceptable IF repetition challenge passed (justified repetition > arbitrary diversity)
Budget Overrun Handling (If >10 agents)
Option 1: Justify Overrun
- Exceptional research complexity: [Explain why >10 agents necessary]
- Critical domain requirements: [Security, medical, financial justification]
- Cost < value: [Why research value justifies budget overrun]
Option 2: Optimize Allocation
- Review dimensions with 3 specialists → Can any drop to 2?
- Consolidate overlapping specialists → Can fact-checker cover verification?
- Reduce total to ≤10 while maintaining coverage
Decision: [Option 1 (justify) or Option 2 (optimize)]
Output: Final validated allocation plan with budget summary, repetition justification, and diversity assessment.
Phase 3e: Decision Logging (Traceability)
After completing Phase 3d, log allocation decisions to project_logs/ for validation and debugging:
Log Files (Both in project_logs/)
File 1: allocation-decision.json (Complete decision trace)
{
"session_id": "[session_id from researchPath]",
"timestamp": "[ISO 8601 timestamp]",
"query": "[original user query]",
"tier": 5,
"methodology": "TODAS",
"phase_1_2_analysis": {
"query_type": "[straightforward/depth-first/breadth-first]",
"novelty": "[low/moderate/high/very high]",
"dimensions": [
{"id": 1, "name": "[name]", "description": "[brief description]"}
]
},
"phase_3a_complexity": [
{
"dimension_id": 1,
"dimension_name": "[name]",
"factors": {
"sub_domains": {"count": 3, "score": 3},
"criticality": {"level": "HIGH", "score": 2},
"novelty": {"level": "MODERATE", "score": 0.5},
"source_diversity": {"types": 3, "score": 3}
},
"total_score": 8.5,
"classification": "COMPLEX",
"recommendation": "2-3 specialists"
}
],
"phase_3b_self_challenge": [
{
"dimension_id": 1,
"initial_selection": "academic-researcher",
"initial_rationale": "[why initially chosen]",
"challenge": {
"alternative": "search-specialist",
"gains": "[what alternative would provide]",
"cons": "[what alternative would lack]",
"comparison": "[initial vs alternative]",
"net_assessment": "[which better matches requirements]",
"verdict": "SWITCH to BOTH / KEEP [initial] / SWITCH to [alternative]",
"decision": "[final decision with reasoning]"
},
"final_selection": ["type1", "type2"],
"final_rationale": "[why final selection optimal]"
}
],
"phase_3c_resource_allocation": [
{
"dimension_id": 1,
"current_allocation": 2,
"coverage_analysis": {
"specialist_coverage": {"specialist1": "40%", "specialist2": "30%"},
"total_coverage": "70%",
"missing": "[what gaps remain]",
"risk": "LOW/MODERATE/HIGH"
},
"options": {
"option_a": {"count": 2, "coverage": "70%", "cost": 5, "risk": "MODERATE"},
"option_b": {"count": 3, "coverage": "95%", "cost": 6, "risk": "LOW"}
},
"decision": "option_a/option_b/option_c",
"final_allocation": 3,
"specialists": ["type1", "type2", "type3"],
"justification": "[why this allocation chosen]"
}
],
"phase_3d_budget_optimization": {
"budget_summary": {
"dimension_allocations": {
"1": {"specialists": 3, "types": ["academic-researcher", "search-specialist", "web-researcher"]},
"2": {"specialists": 1, "types": ["web-researcher"]}
},
"total_specialists": 6,
"fact_checkers": 1,
"grand_total": 7
},
"budget_status": {
"target_range": "5-7",
"maximum": 10,
"actual": 7,
"status": "within_target/above_target/overrun"
},
"repetition_challenge": {
"specialist_type": "web-researcher",
"usage_count": 3,
"dimensions": [1, 2, 4],
"validations": [
{
"dimension_id": 1,
"challenge": "Lazy default or genuine need?",
"fresh_comparison": "[comparison details]",
"verdict": "JUSTIFIED/LAZY_DEFAULT_DETECTED",
"rationale": "[justification]"
}
]
},
"final_diversity": {
"unique_types": 4,
"total_specialists": 6,
"diversity_ratio": 0.67,
"assessment": "HIGH/MODERATE/LOW",
"acceptable": true,
"reason": "[why acceptable/unacceptable]"
}
},
"final_allocation": {
"specialists": [
{"type": "academic-researcher", "dimension": 1, "rationale": "[why chosen for this dimension]"},
{"type": "web-researcher", "dimension": 2, "rationale": "[why chosen]"}
],
"fact_checkers": [
{"dimension": 1, "rationale": "[why fact-checker needed]"}
]
},
"spawned_by": "internet-research-orchestrator",
"version": "1.0-self-challenge"
}
File 2: allocation-decision-summary.json (Quick reference)
{
"session_id": "[session_id from researchPath]",
"timestamp": "[ISO 8601 timestamp]",
"query": "[original user query]",
"tier": 5,
"methodology": "TODAS",
"final_allocation": {
"specialists": [
{"type": "academic-researcher", "dimension": 1, "rationale": "[brief rationale]"},
{"type": "search-specialist", "dimension": 1, "rationale": "[brief rationale]"},
{"type": "web-researcher", "dimension": 1, "rationale": "[brief rationale]"},
{"type": "web-researcher", "dimension": 2, "rationale": "[brief rationale]"},
{"type": "academic-researcher", "dimension": 3, "rationale": "[brief rationale]"},
{"type": "web-researcher", "dimension": 4, "rationale": "[brief rationale]"}
],
"fact_checkers": [
{"dimension": 1, "rationale": "Security dimension verification"}
],
"total_agents": 7
},
"spawned_by": "internet-research-orchestrator"
}
Implementation Instructions
When to log: After Phase 3d completion, BEFORE spawning any agents (Task tool calls)
Location: project_logs/allocation-decisions/[session_folder]/ (session-based organization matching research-sessions pattern)
Directory structure:
project_logs/allocation-decisions/
└── [session_id]/ # e.g., 17112025_115042_mini_app_notification_test3/
├── decision.json # Full decision trace (19KB)
└── summary.json # Quick reference (768B)
Session folder naming: Use session_id from researchPath (format: DDMMYYYY_HHMMSS_topic)
How to write:
session_folder = f"project_logs/allocation-decisions/{session_id}/"
Write(
file_path: f"{session_folder}decision.json",
content: [full JSON with all Phase 3a-3d details]
)
Write(
file_path: f"{session_folder}summary.json",
content: [summary JSON with session_id, timestamp, query, tier, methodology, final_allocation, spawned_by]
)
Benefits:
- ✅ No Data Loss: Each Tier 5 run creates new folder, never overwrites
- ✅ Traceability: Session ID links to research outputs in docs/research-sessions/
- ✅ Validation: Verify self-challenge ran (check phase_3b_self_challenge for verdict changes)
- ✅ Debugging: If test shows web-researcher × 3, check repetition_challenge validations
- ✅ Historical Analysis: Compare allocation decisions across multiple runs
- ✅ Easy Archival: Move entire session folders to archive
- ✅ Quality Metrics: Analyze self-challenge patterns across all sessions
Example validation queries:
jq '.phase_3b_self_challenge[] | select(.challenge.verdict | contains("SWITCH"))' \
project_logs/allocation-decisions/$(ls -t project_logs/allocation-decisions/ | head -1)/decision.json
jq '.phase_3d_budget_optimization.repetition_challenge.validations' \
project_logs/allocation-decisions/17112025_115042_mini_app_notification_test3/decision.json
jq '.final_allocation.specialists | length' \
project_logs/allocation-decisions/17112025_115042_mini_app_notification_test3/summary.json
for session in project_logs/allocation-decisions/*/; do
echo "$(basename $session): $(jq '.final_allocation.total_agents' $session/summary.json) agents"
done
Output: Session folder with both decision files for test validation, debugging, and historical analysis.
Phase 4: Methodical Plan Execution
Execute the plan using adaptive subagent count:
Parallelizable Steps:
- Deploy research-subagent instances using Task tool
- Provide extremely clear task descriptions
- Pass researchPath to ALL subagents (if provided to skill)
- Include tracking parameters (SPAWNED_BY, SESSION_ID, INVOCATION_CONTEXT)
- Spawn in parallel: All Task calls in ONE message (efficiency)
- Wait for completion: Let subagents execute research
- Synthesize findings: Integrate results when complete
Non-Parallelizable/Critical Steps:
- Reasoning-only tasks: Perform yourself (calculations, analysis, formatting)
- Web research tasks: Deploy subagent (orchestrator delegates, not executes)
- Challenging steps: Deploy additional subagents for more perspectives
- Compare results: Use ensemble approach and critical reasoning
Throughout Execution:
- Monitor progress: Continuously check if query being answered
- Update plan: Adapt based on findings from subagents
- Bayesian reasoning: Update priors based on new information
- Adjust depth: If running out of time or diminishing returns, stop spawning and synthesize
- Tactical optimization: Efficiency over completeness when appropriate
Output: Complete research findings from all subagents ready for synthesis.
TodoWrite Integration
Use TodoWrite to track research progress:
Before starting research:
TodoWrite([
{content: "Analyze query and determine type (Phase 1-2)", status: "in_progress", activeForm: "Analyzing query type"},
{content: "Assess dimension complexity scores (Phase 3a)", status: "pending", activeForm: "Assessing dimension complexity"},
{content: "Execute self-challenge phase for specialist selection (Phase 3b)", status: "pending", activeForm: "Executing self-challenge phase"},
{content: "Determine resource allocation counts per dimension (Phase 3c)", status: "pending", activeForm: "Determining resource allocation"},
{content: "Perform budget optimization and repetition challenge (Phase 3d)", status: "pending", activeForm: "Performing budget optimization"},
{content: "Log allocation decisions to project_logs/ (Phase 3e)", status: "pending", activeForm: "Logging allocation decisions"},
{content: "Spawn specialist agents based on final allocation (Phase 4)", status: "pending", activeForm: "Spawning specialist agents"},
{content: "Spawn fact-checker for critical dimension verification (Phase 4)", status: "pending", activeForm: "Spawning fact-checker"},
{content: "Synthesize findings from all specialists (Phase 5)", status: "pending", activeForm": "Synthesizing findings"},
{content: "Report completion with attribution and novelty assessment (Phase 6)", status: "pending", activeForm": "Reporting completion"}
])
As you progress, mark tasks completed and update status. The new Phase 3a-3e steps ensure quality decisions through:
- Complexity assessment (avoid under/over-allocation)
- Self-challenge (catch suboptimal selections)
- Resource allocation (justify specialist counts)
- Repetition challenge (prevent lazy defaults)
- Decision logging (traceability and validation)
Benefits:
- User visibility into research progress
- Clear phase tracking
- Helps avoid skipping steps
Specialist Agent Selection (CRITICAL)
🚨 DO NOT use research-subagent - This is a generic worker type lacking specialized capabilities. You MUST use specialist agents based on research needs.
Available Specialist Agents:
| Agent Type | Use When | Specialized Capabilities |
|---|
| web-researcher | General web queries, current information, broad topics | WebSearch, WebFetch, comprehensive web coverage |
| academic-researcher | Scholarly papers, research publications, scientific topics | Academic databases, peer-reviewed sources, citations |
| search-specialist | Complex queries, deep investigation, hard-to-find info | Boolean operators, advanced search techniques, deep web |
| trend-analyst | Future forecasting, emerging trends, predictions | Weak signal detection, scenario planning, trend analysis |
| market-researcher | Market sizing, segmentation, business intelligence | TAM/SAM/SOM analysis, consumer insights, market data |
| competitive-analyst | Competitor analysis, SWOT, strategic intelligence | Competitive profiling, positioning, industry dynamics |
| synthesis-researcher | Combine findings from multiple sources, meta-analysis | Pattern identification, cross-source integration, synthesis |
| fact-checker | Verify claims, validate sources, check accuracy | Source credibility assessment, claim verification, validation |
Selection Strategy:
- Novel/emerging topics: web-researcher + trend-analyst
- Academic/research topics: academic-researcher + search-specialist
- Market/business topics: market-researcher + competitive-analyst
- Security/compliance: academic-researcher + fact-checker (MANDATORY)
- Multi-source synthesis: synthesis-researcher + fact-checker
Agent Registry Location: .claude/agents/ directory contains all specialist agent definitions.
Subagent Count Guidelines (Adaptive)
TODAS adjusts agent count based on complexity:
| Query Complexity | Subagent Count | Example |
|---|
| Straightforward | 1 specialist | "What is tax deadline this year?" → 1 web-researcher |
| Standard | 2-3 specialists | "Compare top 3 cloud providers" → 3 web-researchers (one per provider) |
| Medium | 3-5 specialists | "Analyze AI impact on healthcare" → 4 agents (academic-researcher, market-researcher, trend-analyst, web-researcher) |
| High | 5-7 specialists | "Fortune 500 CEOs birthplaces/ages" → 7 web-researchers in wave 1, then 7 more (sequential batching) |
Claude Code parallel limit: Maximum 10 parallel tasks, cap at 7 for safety (hooks overhead).
For queries requiring >7 agents: Use sequential batching:
- Spawn 7 specialist agents in parallel
- Wait for completion
- Spawn next 7 specialist agents
- Repeat until coverage complete
Principle: Prefer fewer, more capable specialists over many narrow ones (reduces overhead).
Minimum: Always spawn at least 1 specialist agent for ANY research task (orchestrator delegates, not executes).
Verification Phase: After specialists complete, spawn fact-checker for critical domains (security, compliance, novel topics).
Task Tool Spawning Instructions
How to spawn specialist agent instances:
Step 1: Plan Research Dimensions & Select Specialists
Example: "WebRTC + Web3 convergence in 2025" → 3 dimensions:
- WebRTC current state and 2025 developments → web-researcher (current info)
- Web3 technologies and decentralization trends → trend-analyst (emerging tech)
- Convergence patterns and integration architectures → academic-researcher (research patterns)
Step 2: Call Task Tool for EACH Dimension in ONE Message
🚨 CRITICAL: Use SPECIALIST agent types (web-researcher, academic-researcher, etc.), NOT research-subagent
Correct parallel spawning pattern:
[Call Task tool:]
subagent_type: "web-researcher" ← USE SPECIALIST TYPE
description: "Research WebRTC 2025 developments"
prompt: "Research WebRTC's current state and 2025 roadmap. Focus on:
- Latest WebRTC implementations (2025 features)
- New codec support and performance improvements
- Browser compatibility and adoption trends
Research Path: docs/research-sessions/{session_id}/
SESSION_ID: {session_id}
SPAWNED_BY: internet-research-orchestrator
INVOCATION_CONTEXT: subagent
Output Requirements:
- Save findings to Research Path above
- File naming: webrtc-2025-research-subagent-001.json
- Follow schema: .claude/skills/research/json-schemas/research-output-schema.json
- Size limit: 22K tokens or 20K characters"
[Call Task tool again in SAME message:]
subagent_type: "trend-analyst" ← SPECIALIST for emerging tech
description: "Research Web3 decentralization trends"
prompt: "Research Web3 technologies and decentralization in 2025. Focus on:
- Emerging Web3 protocols (2025)
- Decentralized infrastructure architectures
- Real-world adoption and use cases
[Same tracking parameters and output requirements]"
[Call Task tool again in SAME message:]
subagent_type: "academic-researcher" ← SPECIALIST for patterns/research
description: "Research WebRTC+Web3 convergence patterns"
prompt: "Research integration patterns between WebRTC and Web3. Focus on:
- Decentralized video streaming architectures
- P2P communication with blockchain integration
- Novel use cases emerging in 2025
[Same tracking parameters and output requirements]"
All Task calls in ONE message = parallel execution (Claude Code optimization).
Step 3: Spawn Fact-Checker (After Specialists Complete)
For critical domains (security, compliance, novel topics), spawn fact-checker:
[After specialists complete, call Task tool:]
subagent_type: "fact-checker"
description: "Verify critical claims from research"
prompt: "Verify key claims from WebRTC, Web3, and convergence research. Check:
- Citation accuracy and source credibility
- Conflicting information across sources
- Confidence levels for novel 2025 claims
- Gaps in coverage or missing perspectives
Research Path: docs/research-sessions/{session_id}/
SESSION_ID: {session_id}
SPAWNED_BY: internet-research-orchestrator
Target verification rate: >75% (good), >85% (excellent)"
What NOT to Do
❌ Using research-subagent (lacks specialized capabilities):
Task(subagent_type="research-subagent", ...) ← WRONG - use specialist agents
❌ Sequential spawning (slow):
Task(...), wait for completion, then Task(...), wait, then Task(...)
❌ Bash workarounds (breaks hooks integration):
Bash: claude_cli task run research-subagent "$(cat /tmp/task.md)" &
❌ Creating task files and stopping (NOT spawning):
Write task description to .txt file, then don't call Task tool
❌ Skipping fact-checker for critical domains (security, compliance, novel topics):
[Spawn specialists, synthesize, report] ← Missing verification phase
What to ALWAYS Do
✅ Use SPECIALIST agent types (web-researcher, academic-researcher, trend-analyst, etc.)
✅ Use Task tool (it's available to you)
✅ Call Task multiple times in ONE message (parallel spawning)
✅ Spawn specialists IMMEDIATELY after planning (efficiency)
✅ Spawn fact-checker for critical domains (security, compliance, novel topics)
✅ Pass researchPath to ALL agents (when provided to skill)
✅ Include tracking parameters (SPAWNED_BY, SESSION_ID, INVOCATION_CONTEXT)
Delegation Rules (CRITICAL)
Main Claude delegates ALL research to subagents:
Core Orchestration Principles
-
Orchestrator role: You coordinate and synthesize, NOT execute primary research
- Spawn 1-7 research-subagent instances (adaptive based on complexity)
- Provide each subagent with extremely detailed, specific instructions
- Let subagents perform all web searches, fact-finding, and information gathering
- Focus on planning, analyzing, integrating findings, identifying gaps
-
Verification requirement: After planning, count how many subagents you plan to spawn
- If count = 0, revise plan immediately
- Every query requires AT LEAST 1 research-subagent
- Your value is orchestration and synthesis, not execution
-
ResearchPath coordination: When invoked with researchPath parameter
- ALL subagents MUST save outputs to SAME researchPath
- Pass researchPath to EVERY subagent you spawn
- File coordination is MANDATORY for multi-agent research
Subagent Task Descriptions
Provide each subagent with:
- Specific research objective: Ideally 1 core objective per subagent
- Expected output format: List, report, answer, analysis, etc.
- Background context: How subagent contributes to overall research plan
- Key questions: What to answer as part of research
- Starting points and sources: Define reliable information, list unreliable sources to avoid
- Specific tools: WebSearch, WebFetch for internet information gathering
- Scope boundaries: Prevent research drift (if needed)
- Output requirements: When researchPath provided (see section above)
Validation: If all subagents follow instructions well, aggregate results should allow EXCELLENT answer to user query (complete, thorough, detailed, accurate).
Deployment Strategy
Priority and dependency:
- Deploy most important subagents first
- If tasks depend on results from specific task, create that blocking subagent first
- Ensure sufficient coverage for comprehensive research
- All substantial information gathering delegated to subagents
Avoid overlap:
- Every subagent should have distinct, clearly separate tasks
- Prevent replicating work unnecessarily
- Avoid wasting resources on redundant research
Efficiency while waiting:
- Analyze previous results
- Update research plan
- Reason about user's query and how to best answer it
- Do NOT idle waiting for subagents
Response Instructions
Before Providing Final Answer
- Review most recent facts compiled during research process
- Reflect deeply: Can these facts answer query sufficiently?
- Provide final answer in format best for user's query
- When invoked with researchPath: Return summary of research completed and file locations (research skill handles synthesis)
- When invoked standalone: Format final research report in Markdown with proper source attribution
Source Attribution (Tier 5 Critical)
For novel/emerging domains, transparency is essential:
- Inline references: "According to [Source Name]", "Research from [Organization] shows..."
- Sources section: At end of report, list all key sources consulted
- Credibility: Builds trust and allows verification
- Recency: Note publication dates (critical for 2025+ topics)
Novelty Assessment
Include in synthesis:
- Novelty level: Is this truly emerging? Post-training? Unprecedented?
- Confidence: How much information available? (Emerging topics = less data)
- Gaps: What's unknown? What requires future research?
- Verification challenges: Difficult to verify emerging claims (note limitations)
Examples
Example 1: Straightforward Novel Query (1 Subagent)
User query: "What is Claude Sonnet 4.5 and when was it released?"
Main Claude reasoning:
- Query type: Straightforward (single focused investigation)
- Novelty: High (Sonnet 4.5 released post-training)
- Subagent count: 1 (simple lookup)
TodoWrite tracking:
[
{content: "Determine query type: Straightforward", status: "completed"},
{content: "Spawn 1 research-subagent for Sonnet 4.5 info", status: "in_progress"},
{content: "Synthesize findings and report", status: "pending"}
]
Task tool call:
Task(
subagent_type: "research-subagent",
description: "Research Claude Sonnet 4.5 release",
prompt: "Research Claude Sonnet 4.5 model. Find:
- Official release date
- Model capabilities and improvements
- Comparison to previous Sonnet versions
- Anthropic's official announcements
Sources: Prioritize Anthropic official site, reputable tech news
SESSION_ID: {session_id}
SPAWNED_BY: internet-research-orchestrator"
)
Synthesis: After subagent returns, Main Claude synthesizes findings with source attribution and reports completion.
Example 2: Depth-First Novel Query (3-5 Subagents)
User query: "Analyze multimodal AI agent frameworks emerging in 2025 - what are the leading approaches and their trade-offs?"
Main Claude reasoning:
- Query type: Depth-first (multiple perspectives on same issue)
- Novelty: Very high (2025 developments, post-training)
- Subagent count: 4 (different approaches/perspectives)
Research plan:
- Subagent 1: Tool-augmented agents (AutoGPT, LangChain evolution)
- Subagent 2: Reasoning-first agents (Chain-of-Thought, ReAct patterns)
- Subagent 3: Vision-language agents (GPT-4V descendants, Gemini Ultra)
- Subagent 4: Code-execution agents (Code Interpreter evolution)
TodoWrite tracking:
[
{content: "Determine query type: Depth-first", status: "completed"},
{content: "Develop 4-perspective research plan", status: "completed"},
{content: "Spawn 4 research-subagents in parallel", status: "in_progress"},
{content: "Synthesize findings with trade-off analysis", status: "pending"}
]
Task tool calls (all in ONE message for parallel execution):
Task(subagent_type: "research-subagent", description: "Research tool-augmented agents", prompt: "[Detailed instructions for tool-augmented approach research]")
Task(subagent_type: "research-subagent", description: "Research reasoning-first agents", prompt: "[Detailed instructions for reasoning patterns research]")
Task(subagent_type: "research-subagent", description: "Research vision-language agents", prompt: "[Detailed instructions for multimodal capabilities research]")
Task(subagent_type: "research-subagent", description: "Research code-execution agents", prompt: "[Detailed instructions for code generation research]")
Synthesis: After all 4 subagents return, Main Claude integrates findings, identifies trade-offs across approaches, assesses novelty and emerging trends, provides source-attributed analysis.
Example 3: Breadth-First Novel Query (5-7 Subagents)
User query: "Compare the top 5 decentralized video streaming platforms in 2025 across performance, cost, adoption, and technology stack."
Main Claude reasoning:
- Query type: Breadth-first (5 independent platform researches)
- Novelty: High (2025 platforms, emerging Web3 space)
- Subagent count: 5 (one per platform)
Research plan:
- Identify top 5 decentralized video platforms (preliminary research)
- For each platform, research: performance, cost, adoption, tech stack
- Spawn 5 parallel subagents (one per platform)
- Aggregate findings into comparison table
TodoWrite tracking:
[
{content: "Identify top 5 decentralized video platforms", status: "completed"},
{content: "Spawn 5 research-subagents (one per platform)", status: "in_progress"},
{content: "Aggregate findings into comparison analysis", status: "pending"}
]
Task tool calls (all in ONE message):
Task(subagent_type: "research-subagent", description: "Research Platform A", prompt: "[Detailed instructions for Platform A across 4 dimensions]")
Task(subagent_type: "research-subagent", description: "Research Platform B", prompt: "[...]")
Task(subagent_type: "research-subagent", description: "Research Platform C", prompt: "[...]")
Task(subagent_type: "research-subagent", description: "Research Platform D", prompt: "[...]")
Task(subagent_type: "research-subagent", description: "Research Platform E", prompt: "[...]")
Synthesis: After all 5 subagents return, Main Claude creates comparison table, identifies patterns across platforms, assesses maturity and adoption trends, provides source-attributed recommendations.
Example 4: Multi-Dimensional Query with Self-Challenge & Resource Allocation (5-7 Specialists)
User query: "Research emerging approaches for secure push notifications in multi-tenant mini-app platforms for 2025, covering security architecture, infrastructure scalability, real-time coordination, and mobile-native implementations."
Phase 1-2: Query Analysis & Dimension Breakdown
Query type: Breadth-first (4 distinct dimensions)
Novelty: Very high (2025 emerging approaches, post-training)
Dimensions identified:
- Security architecture (multi-tenant isolation, encryption)
- Infrastructure scalability (cloud, edge, hybrid)
- Real-time coordination (WebSocket, MQTT, emerging protocols)
- Mobile-native implementations (iOS/Android push mechanisms)
Phase 3a: Dimension Complexity Assessment
Dimension 1: Security Architecture
- Sub-domains: 3 (isolation, encryption, compliance) → +3
- Criticality: HIGH (security domain) → +2
- Novelty: MODERATE (2025 approaches) → +0.5
- Source diversity: 3 types (papers, standards, blogs) → +3
- Complexity Score: 8.5 (COMPLEX)
Dimension 2: Infrastructure Scalability
- Sub-domains: 2 (cloud, edge) → +2
- Criticality: MODERATE → +1
- Novelty: MODERATE → +0.5
- Source diversity: 2 types (vendor docs, case studies) → +2
- Complexity Score: 5.5 (MODERATE)
Dimension 3: Real-Time Coordination
- Sub-domains: 2 (protocols, patterns) → +2
- Criticality: MODERATE → +1
- Novelty: HIGH (emerging 2025 protocols) → +1
- Source diversity: 2 types (research, blogs) → +2
- Complexity Score: 6 (MODERATE)
Dimension 4: Mobile-Native Implementations
- Sub-domains: 1 (push APIs) → +1
- Criticality: LOW → +0
- Novelty: LOW (established APIs) → +0
- Source diversity: 1 type (platform docs) → +2
- Complexity Score: 3 (SIMPLE)
Phase 3b: Self-Challenge Phase (Specialist Selection)
Dimension 1: Security Architecture (score 8.5 = COMPLEX)
Initial Selection: academic-researcher
- Rationale: Security research papers, theoretical frameworks, cryptography studies
🚨 CHALLENGE: "Wait, why didn't I choose search-specialist instead?"
Alternative: search-specialist
- Gains if chosen: Deep investigation, compliance standards (ISO 27001, SOC 2), hard-to-find threat reports
- Cons if chosen: Less academic rigor, may miss theoretical security frameworks
- Comparison: academic-researcher (theory + papers) vs search-specialist (standards + compliance)
- Net assessment: Dimension needs BOTH theory AND compliance
- Verdict: ❌ SWITCH to BOTH → Upgrade to 2 specialists (academic-researcher + search-specialist)
Dimension 2: Infrastructure Scalability (score 5.5 = MODERATE)
Initial Selection: web-researcher
- Rationale: Cloud provider docs (AWS, GCP, Azure), infrastructure engineering blogs
🚨 CHALLENGE: "Wait, why didn't I choose market-researcher instead?"
Alternative: market-researcher
- Gains: Market sizing, vendor comparisons, adoption trends
- Cons: Less technical depth, focus on business metrics vs technical scalability
- Comparison: web-researcher (technical docs) vs market-researcher (business analysis)
- Net: Query asks "scalability approaches" (technical), not "market landscape" (business)
- Verdict: ✅ KEEP web-researcher (technical focus matches dimension requirements)
Dimension 3: Real-Time Coordination (score 6 = MODERATE)
Initial Selection: web-researcher
- Rationale: WebSocket docs, MQTT protocols, real-time coordination blogs
🚨 CHALLENGE: "Wait, why didn't I choose academic-researcher instead?"
Alternative: academic-researcher
- Gains: Coordination algorithms, distributed systems research papers, emerging 2025 academic protocols
- Cons: Less practical implementation guidance
- Comparison: web-researcher (protocols + docs) vs academic-researcher (algorithms + research)
- Net: Dimension includes "emerging 2025 approaches" → academic research valuable
- Verdict: ❌ SWITCH to academic-researcher (self-challenge caught "emerging 2025" keyword mismatch)
Dimension 4: Mobile-Native (score 3 = SIMPLE)
Initial Selection: web-researcher
- Rationale: Apple/Google platform docs, push notification guides
- Verdict: ✅ KEEP (simple dimension, platform docs sufficient)
Phase 3c: Resource Allocation Challenge
Dimension 1: Security Architecture (COMPLEX, score 8.5)
Current Allocation: 1 specialist (academic-researcher → revised to academic-researcher + search-specialist in Phase 3b)
🚨 CHALLENGE: "Should this dimension get MORE than 2 specialists?"
Coverage Analysis (with 2 specialists):
- academic-researcher: Security theory, cryptography papers (40%)
- search-specialist: Compliance standards, threat landscape (30%)
- Total Coverage: 70%
- Missing: Current 2025 threat intelligence, industry security reports (30%)
- Risk: ⚠️ MODERATE - Security domain, 70% may be insufficient
Option C: Add web-researcher (3 specialists total)
- Additional: web-researcher covering 2025 CVEs, industry security blogs, vendor security docs
- Coverage: 70% → 95%
- Cost: +1 agent (total 6)
- Risk: ✅ LOW
- Decision: ✅ UPGRADE to 3 specialists (security criticality justifies comprehensive coverage)
Dimension 2, 3, 4: All remain at 1 specialist (coverage >80%, moderate/simple complexity)
Phase 3d: Budget Optimization & Final Verification
Budget Summary:
- Dimension 1 (Security): 3 specialists (academic-researcher, search-specialist, web-researcher)
- Dimension 2 (Infrastructure): 1 specialist (web-researcher)
- Dimension 3 (Real-Time): 1 specialist (academic-researcher)
- Dimension 4 (Mobile-Native): 1 specialist (web-researcher)
- Total Specialists: 6
- Fact-Checkers: 1 (for security dimension verification)
- Grand Total: 7 agents
Budget Status: ⚠️ Within Max (7 agents, target 5-7) ✅
Repetition Challenge:
Specialist Type: web-researcher (used 3 times)
Used For: Dimension 1 (Security - 3rd specialist), Dimension 2 (Infrastructure), Dimension 4 (Mobile-Native)
Dimension 1 Security - web-researcher (3rd specialist):
- 🚨 Challenge: "Lazy default or genuine need?"
- Fresh comparison: web-researcher (2025 CVEs, blogs) vs trend-analyst (future threats)
- Net: Query asks "2025 approaches" (current + emerging), not "2026 forecast"
- Verdict: ✅ JUSTIFIED - 2025 current threat intelligence required
Dimension 2 Infrastructure - web-researcher:
- 🚨 Challenge: Already using for Dim 1, genuine here?
- Fresh comparison: web-researcher (cloud docs) vs market-researcher (vendor comparison)
- Net: Technical scalability > vendor comparison
- Verdict: ✅ JUSTIFIED - Technical dimension match
Dimension 4 Mobile-Native - web-researcher:
- 🚨 Challenge: Third usage, lazy default?
- Fresh comparison: web-researcher (platform docs) vs academic-researcher (push theory)
- Net: Production implementation focus
- Verdict: ✅ JUSTIFIED - Platform docs optimal for mobile-native
Final Diversity: 4 unique types / 6 specialists = 67% (MODERATE) ✅ Acceptable (all repetitions justified)
Phase 3e: Decision Logging
Before spawning agents, log decisions to project_logs/:
allocation-decision.json (full trace):
{
"session_id": "17112025_112316_mini_app_notification",
"timestamp": "2025-11-17T11:23:16Z",
"query": "Research emerging approaches for secure push notifications in multi-tenant mini-app platforms for 2025...",
"tier": 5,
"methodology": "TODAS",
"phase_3a_complexity": [
{"dimension_id": 1, "dimension_name": "Security Architecture", "total_score": 8.5, "classification": "COMPLEX"},
{"dimension_id": 2, "dimension_name": "Infrastructure Scalability", "total_score": 5.5, "classification": "MODERATE"},
{"dimension_id": 3, "dimension_name": "Real-Time Coordination", "total_score": 6, "classification": "MODERATE"},
{"dimension_id": 4, "dimension_name": "Mobile-Native Implementations", "total_score": 3, "classification": "SIMPLE"}
],
"phase_3b_self_challenge": [
{"dimension_id": 3, "initial_selection": "web-researcher", "challenge": {"verdict": "SWITCH to academic-researcher"}, "final_selection": ["academic-researcher"]}
],
"phase_3c_resource_allocation": [
{"dimension_id": 1, "final_allocation": 3, "specialists": ["academic-researcher", "search-specialist", "web-researcher"]}
],
"phase_3d_budget_optimization": {
"budget_summary": {"total_specialists": 6, "fact_checkers": 1, "grand_total": 7},
"repetition_challenge": {"specialist_type": "web-researcher", "usage_count": 3, "all_justified": true}
},
"spawned_by": "internet-research-orchestrator"
}
allocation-decision-summary.json (quick reference):
{
"session_id": "17112025_112316_mini_app_notification",
"timestamp": "2025-11-17T11:23:16Z",
"query": "Research emerging approaches for secure push notifications...",
"tier": 5,
"methodology": "TODAS",
"final_allocation": {
"specialists": [
{"type": "academic-researcher", "dimension": 1, "rationale": "Security theory, cryptography papers"},
{"type": "search-specialist", "dimension": 1, "rationale": "Compliance standards"},
{"type": "web-researcher", "dimension": 1, "rationale": "2025 threat intelligence"},
{"type": "web-researcher", "dimension": 2, "rationale": "Cloud infrastructure docs"},
{"type": "academic-researcher", "dimension": 3, "rationale": "Emerging 2025 coordination algorithms"},
{"type": "web-researcher", "dimension": 4, "rationale": "Mobile platform documentation"}
],
"fact_checkers": [{"dimension": 1, "rationale": "Security verification"}],
"total_agents": 7
},
"spawned_by": "internet-research-orchestrator"
}
TodoWrite tracking:
[
{content: "Analyze query and determine type (Phase 1-2)", status: "completed"},
{content: "Assess dimension complexity scores (Phase 3a)", status: "completed"},
{content: "Execute self-challenge phase for specialist selection (Phase 3b)", status: "completed"},
{content: "Determine resource allocation counts per dimension (Phase 3c)", status: "completed"},
{content: "Perform budget optimization and repetition challenge (Phase 3d)", status: "completed"},
{content: "Log allocation decisions to project_logs/ (Phase 3e)", status: "completed"},
{content: "Spawn 6 specialist agents based on final allocation (Phase 4)", status: "in_progress"},
{content: "Spawn 1 fact-checker for security verification (Phase 4)", status: "pending"},
{content: "Synthesize findings from all 7 agents (Phase 5)", status: "pending"}
]
Task tool calls (Phase 4 - all in ONE message for parallel execution):
Task(subagent_type: "academic-researcher", description: "Research security architecture theory", prompt: "[Security papers, cryptography, isolation frameworks]")
Task(subagent_type: "search-specialist", description: "Research security compliance standards", prompt: "[ISO 27001, SOC 2, GDPR for multi-tenant]")
Task(subagent_type: "web-researcher", description: "Research 2025 security threat intelligence", prompt: "[CVEs, security blogs, vendor advisories]")
Task(subagent_type: "web-researcher", description: "Research infrastructure scalability", prompt: "[Cloud providers, edge computing, hybrid architectures]")
Task(subagent_type: "academic-researcher", description: "Research real-time coordination algorithms", prompt: "[Emerging 2025 protocols, distributed coordination papers]")
Task(subagent_type: "web-researcher", description: "Research mobile-native push implementations", prompt: "[iOS APNs, Android FCM, platform best practices]")
Task(subagent_type: "fact-checker", description: "Verify security dimension claims", prompt: "[Validate Dimension 1 findings for accuracy]")
Key Decisions Demonstrated:
- ✅ Self-challenge caught suboptimal selection (Dimension 3: web-researcher → academic-researcher)
- ✅ Resource allocation upgraded critical dimension (Dimension 1: 1 → 3 specialists)
- ✅ Repetition challenge validated all web-researcher × 3 usage (not lazy defaults)
- ✅ Budget within target (7 agents, justified by security criticality)
- ✅ Final allocation: 6 specialists + 1 fact-checker = comprehensive multi-dimensional coverage
Key Differences from Other Tiers
Tier 5 (TODAS) vs Tier 4 (RBMAS):
- Adaptive agent count: 1-7 (vs fixed 3-7)
- Tactical optimization: Stop at diminishing returns (vs complete phases)
- Novel domain focus: Post-training, emerging (vs established domains)
- Query type classification: Depth/Breadth/Straightforward (vs dimension counting)
- Synthesis approach: Novelty assessment, uncertainty quantification (vs cross-dimensional patterns)
Tier 5 (TODAS) vs Tier 3 (Light Parallel):
- Complexity: Novel/emerging domains (vs standard 2-4 dimensions)
- Agent range: 1-7 adaptive (vs 2-4 fixed)
- Methodology: TODAS 4-phase (vs simple parallel coordination)
- Synthesis: Novelty assessment (vs aggregation)
When to use Tier 5 vs others:
- ✅ Tier 5: Novel domains, post-training data, emerging tech, unprecedented topics
- ✅ Tier 4: Established domains, comprehensive multi-dimensional research (4+ dimensions)
- ✅ Tier 3: Standard research, 2-4 dimensions, known patterns
- ✅ Tier 1-2: Simple lookups, specialist queries
Skill Type: Research Orchestration (Tier 5 - Novel Domains)
Methodology: TODAS (Tactical Optimization & Depth-Adaptive System)
Agent Range: 1-7 subagents (adaptive based on complexity)
Specialization: Novel/emerging domains, post-training information, unprecedented topics
Created: Phase 4 of agent-to-skill migration
Version: 1.0
