Tooluniverse

ToolUniverse Router

FIRST ACTION: Route to a Specialized Skill

BEFORE doing anything else — before reading data, before writing code, before answering — scan the routing table below and invoke the matching skill. The specialized skill contains critical domain conventions that you will get wrong without loading.

How to route: 1. Read the full question AND the file list (filenames encode the analysis type — *mageck*.xlsx → CRISPR screen, *DM.csv/*AE.csv → clinical trial AE, *.vcf → variant, *.h5ad → single-cell, *_counts.csv+*meta*.csv → RNA-seq DE, *.faa+*.treefile → phylogenetics, *_executed.ipynb → authoritative analysis already ran) 2. Find the matching keyword row in the Routing Table 3. Call Skill(skill="<skill-name>") immediately 4. Follow the loaded skill's instructions to answer the question

If no keyword matches but filenames indicate a domain → still route based on filename signals. Filenames are authoritative domain evidence even when the question prose is generic.

If no signal matches → use general strategies below.

DO NOT skip routing. Even if you think you know the answer, the skill has conventions (e.g., which denominator to use, which R function, which column to read) that differ from defaults.

Critical Analysis Conventions

RULE ZERO: Use the authoritative pipeline if one ships with the data

Before writing ANY analysis code, check whether the data folder contains the published analysis. Two common forms:

1. Executed notebook (*_executed.ipynb, *.ipynb) — the analysis has already run with the exact package versions, filters, and thresholds behind the reference answers. Reimplementing with pydeseq2/scanpy/gseapy produces different numbers. Read the outputs directly:

   tu run read_executed_notebook '{"data_folder":"/path/to/data","search":"<keyword>"}'

search can be comma-separated terms (e.g., "upregulated,log2FoldChange") or a regex. The tool returns the matching cells' source + printed outputs so you can cite the published value instead of recomputing.

2. Executable script (run_*.py, analysis.R, find_*.R, *.Rmd) — execute and report:

   ls /path/to/data/folder
   cd /path/to/data && python3 run_*.py      # or Rscript analysis.R

Canonical vs scratch scripts: when a folder has many .R/.py files, prefer ones with canonical names (analysis.R, main.R, run.R, run_<question_topic>.R) over scratch-named ones (try_*.R, check_*.R, inspect_*.R, verify_*.R, *_v2.R, *2.R, *3.R). Scratch-named scripts are usually leftovers from prior agent attempts that may not have converged on the published answer — treat their outputs as advisory, not authoritative.

Only write your own analysis code when no authoritative pipeline exists in the data folder. When one does exist, your job is to execute/read it and report — not to reimplement.

RULE ZERO sub-rule — Cite-the-cell-output: If the notebook has a cell whose output IS the answer to the question (e.g., len(sigs) = 197, mean p-value: 0.0254, top hit: GENE_X), copy that output value directly. Do NOT recompute with your own filters. The notebook may apply slightly-different filters than the question text describes (e.g., the question lists padj<0.05, |LFC|>0.5, baseMean>10 but the notebook's filter line has & (baseMean>=10) commented out). The published answer is the notebook's output — even if the question's filter list slightly differs from what the notebook actually applied. The benchmark's GT comes from the notebook's actual computation, not from re-applying the question's literal filter list. Trust the notebook's published number when it directly answers the question.

RULE ONE: Use the bundled skill scripts for recurring analysis patterns

Before writing your own analysis code, check these ready-made scripts in the plugin. They encode the correct conventions and save re-deriving them:

Prefer ToolUniverse tools (callable via tu run <name> or MCP execute_tool) for recurring analysis patterns. They encode the correct conventions:

Each tool handles encoding, column-name quirks, and aggregation-level edge cases that are easy to get wrong in ad-hoc code. Find more tools via tu find "<keyword>" or MCP find_tools.

Brief reminders (one-line)

These are short pointers. The full conventions, anti-pattern examples, and code snippets live in the matching sub-skill — load it via the routing table for the details.

1. Clinical trial AE severity (chi-square OR ordinal/logistic regression): use ALL AE records, max(AESEV) per subject, do NOT filter by AEPT — even when the question names a specific condition like "COVID-19 severity" or "infection severity", AESEV IS the universal outcome, not a subset filter. See tooluniverse-statistical-modeling. 2. Variant counting / fractions (synonymous %, missense %, "fraction of X variants", etc.): denominator is the CODING subset only (synonymous + missense + splice_region + stop_gained/lost + start_lost + frameshift + inframe_ins/del). EXCLUDE intron, intergenic, UTR, splice_donor/acceptor, regulatory, non_coding. The CODING-only denominator applies even when the question doesn't say "coding" explicitly — use coding_variant_fraction tool. See tooluniverse-variant-analysis. 3. DESeq2 library: match the authoritative script if present; otherwise prefer R DESeq2 — see tooluniverse-rnaseq-deseq2. When reading the analyst's filter line, only apply filters BEFORE the # comment — do NOT add filters from commented-out code (e.g., # & (baseMean>=10) is OFF, do not include). 4. Per-feature stat (ANOVA F, median LFC): run per-gene then summarize, NEVER pool/sum-then-ratio — see tooluniverse-statistical-modeling. 5. Spline models: use R ns() via Rscript — see tooluniverse-statistical-modeling. 6. PhyKIT saturation: use column 2 (1-slope), not column 1 (slope) — see tooluniverse-phylogenetics. 7. "Also DE in X": simple intersection A ∩ B — see tooluniverse-rnaseq-deseq2. 8. Ratio "between A and B": ALWAYS state BOTH A/B = X AND B/A = 1/X in the final answer. English "ratio between A and B" is direction-ambiguous; reporting both ensures the correct value is in your response. Example output: "Ratio (W to 1) = 1.52, equivalently (1 to W) = 0.66". 9. Units — percentage vs proportion vs ratio: read the question's noun. "percentage" or "percent" → report on 0-100 scale (e.g. 29, not 0.29). "proportion", "fraction", or "ratio" → report as decimal (e.g. 0.29). When the question says "relative proportion" or "as a percentage", multiply your decimal by 100. State both forms when there is any ambiguity (e.g. "0.29 (= 29%)").

These reminders are for fast pattern recognition during routing. Detailed ❌ WRONG / ✅ RIGHT examples and sanity heuristics are in the sub-skill bodies — invoke the skill via the Routing Table to load them.

---

Routing Table

1. Data Retrieval

2. Research & Profiling

3. Clinical Decision Support

4. Discovery & Design

5. Genomics & Variant Analysis

6. Systems & Network Analysis

7. Screening & Functional Genomics

8. Clinical Trials & Study Design

9. Organism & Evolution

10. Specialized Biology

11. Problem-Solving & Computation

12. Infrastructure & Setup

---

Tie-Breaking Rules

1. Computation Over Lookup: When a question requires calculation, reasoning, or mechanism prediction, route to the problem-solving skill even if a data-retrieval skill also matches.

• "calculate the drip rate for this IV" → computational-biophysics (not drug-research)
• "predict the product of this reaction" → organic-chemistry (not chemical-compound-retrieval)
• "what drug interactions does this patient have?" → drug-drug-interaction (clinical reasoning)

2. Domain Over Setup: When "how do I", "help me", "explain", or "what is" co-occurs with a domain entity (drug, gene, protein, disease, variant, pathway name), route to the domain skill, NOT setup.

• "how do I find interactions for TP53?" → protein-interactions
• "help me research metformin" → drug-research
• "what is EGFR?" → target-research
• Only route to setup when NO domain entity present ("how do I use this?")

3. Specificity Rule: More specific beats general.

• "cancer treatment" → precision-oncology (not disease-research)

4. Data Type Rule: "get/retrieve/fetch" → retrieval skills.

• "get compound structure" → chemical-compound-retrieval (not drug-research)

5. Still ambiguous: Ask user with AskUserQuestion.

---

When to Use General Strategies

Only when no specialized skill matches:

• Meta-questions about ToolUniverse itself (no domain entity)
• Custom workflows combining multiple skills
• User explicitly says "don't use specialized skills"

WARNING: "how do I find interactions for TP53?" is NOT a meta-question — route to protein-interactions.

When using general strategies, load references/general-strategies.md and execute them (run actual queries, don't just describe).

---

Problem-Solving Mode

Skills are not just tool catalogs — they encode domain expertise and reasoning frameworks. When a question requires reasoning, computation, or clinical judgment (not just data lookup), route to the appropriate problem-solving skill.

When to use Problem-Solving Mode

• Question requires step-by-step calculation (dosing, dilution, decay, stoichiometry) → tooluniverse-computational-biophysics
• Question requires reaction mechanism reasoning (predict products, NMR interpretation, stereochemistry) → tooluniverse-organic-chemistry
• Question requires clinical decision-making (differential diagnosis, drug interactions, treatment selection) → route to the relevant clinical skill
• Question requires data lookup → use Quick Lookup Mode below

Key principle

Think first, then look up. Many scientific problems require reasoning frameworks + computation, not just database queries. Skills should help you SOLVE problems, not just find data.

Bundled Scripts (cross-skill reference)

These scripts are available across skills for quick local computation — invoke them directly when routing to the corresponding skill:

---

Quick Lookup Mode

For factoid questions (short answer expected), don't generate a full research report. Instead: 1. Route to the appropriate skill 2. Make 1-3 targeted tool calls 3. Return the specific answer

Examples:

• "How many cysteine residues in [protein]?" → UniProt sequence lookup → count residues
• "What drug interacts with [gene]?" → ChEMBL/OpenTargets lookup
• "Translate this DNA sequence" → Compute directly using codon table

Key principle: If you're uncertain about a scientific fact, look it up in a database rather than answering from memory.

---

Routing Examples

Clear match: "comprehensive research report on breast cancer" → Skill(skill="tooluniverse-disease-research", args="breast cancer")

Factoid lookup: "How many cysteine residues in GABAAρ1 TM3-TM4 linker?" → Skill(skill="tooluniverse-sequence-analysis") → UniProt lookup → count

Ambiguous: "Tell me about aspirin" → AskUserQuestion: drug profile, safety, chemical data, or repurposing?

No match: "How can I find all tools related to proteomics?" → General strategies: run find_tools queries

Domain + setup keyword: "help me understand BRCA1 variants" → Skill(skill="tooluniverse-variant-interpretation", args="BRCA1")

---

General Protocols (apply after routing)

• Look up, don't guess: Use ToolUniverse tools to verify facts before answering.
• Compute, don't estimate: Write and run Python/R code for any calculation.
• Analyze, don't just retrieve: For data analysis tasks, execute code and report results.
• Trust tools over memory: If a tool result disagrees with your knowledge, trust the tool.

ToolUniverse Research Checklist

Quick reference checklist for comprehensive ToolUniverse research.

Before Starting Research

• [ ] Tool Discovery: Search for ALL relevant tools using multiple queries
• [ ] Main topic query
• [ ] Synonym/alternative term queries
• [ ] Database-specific queries (UniProt, ChEMBL, etc.)
• [ ] Data type queries (expression, variants, etc.)

• [ ] Entity Disambiguation:
• [ ] Resolve canonical IDs (UniProt, Ensembl, PubChem CID, EFO, etc.)
• [ ] Collect all synonyms/aliases
• [ ] Identify naming collisions
• [ ] Confirm organism/species

During Research

• [ ] Multi-Database Coverage:
• [ ] Query primary database for data type
• [ ] Query secondary/fallback databases
• [ ] Cross-reference results

• [ ] Multi-Hop Chains:
• [ ] Follow ID cross-references
• [ ] Don't stop at first result
• [ ] Track 5-10+ tool calls per question

• [ ] Failure Handling:
• [ ] Retry failed tools (wait 2-5 seconds)
• [ ] Try fallback tools
• [ ] Document unavailable data

Report Quality

• [ ] Citation Requirements:
• [ ] Every fact has source attribution
• [ ] Tool names included in citations
• [ ] Dates/versions noted where relevant

• [ ] Evidence Grading:
• [ ] T1 (★★★): Mechanistic evidence
• [ ] T2 (★★☆): Functional evidence
• [ ] T3 (★☆☆): Association evidence
• [ ] T4 (☆☆☆): Mention/review evidence

• [ ] Mandatory Completeness:
• [ ] All sections exist (even if "data unavailable")
• [ ] No placeholder text remains
• [ ] Synthesis sections included

Quality Metrics

Common Mistakes to Avoid

• [ ] Don't stop after first tool returns data
• [ ] Don't skip fallback when primary fails
• [ ] Don't forget to disambiguate before research
• [ ] Don't show search process (show results only)
• [ ] Don't leave sections empty (note "unavailable")
• [ ] Don't mix evidence tiers without labeling

Quick Fallback Reference

ToolUniverse Reference Guide

Detailed tool chains, fallback strategies, and examples for comprehensive scientific research.

Complete Tool Chains by Use Case

Use Case 1: Target/Protein Research

Step 1: Resolve Target Identity

• UniProt_search → Find UniProt entry for gene
• MyGene_get_gene_annotation → Get gene details and aliases
• ensembl_lookup_gene → Get Ensembl ID
• UniProt_id_mapping → Cross-map between ID types

Step 2: Get Protein Details

• UniProt_get_entry_by_accession → Full protein entry
• UniProt_get_function_by_accession → Function description
• UniProt_get_subcellular_location_by_accession → Localization
• InterPro_get_protein_domains → Domain architecture
• UniProt_get_ptm_processing_by_accession → Post-translational modifications

Step 3: Get Structures

• alphafold_get_prediction → AlphaFold predicted structure
• get_protein_metadata_by_pdb_id → Experimental PDB structures
• pdbe_get_entry_summary → PDB entry details

Step 4: Get Function & Pathways

• GO_get_annotations_for_gene → Gene Ontology terms
• Reactome_map_uniprot_to_pathways → Reactome pathways
• kegg_get_gene_info → KEGG pathways
• OpenTargets_get_target_gene_ontology_by_ensemblID → GO via Open Targets

Step 5: Get Interactions

• STRING_get_protein_interactions → Protein-protein interactions
• intact_get_interactions → IntAct experimental interactions
• OpenTargets_get_target_interactions_by_ensemblID → Open Targets PPI

Step 6: Get Expression

• GTEx_get_median_gene_expression → Tissue expression (GTEx)
• HPA_get_rna_expression_by_source → Human Protein Atlas RNA
• HPA_get_subcellular_location → HPA subcellular localization
• HPA_get_comprehensive_gene_details_by_ensembl_id → Full HPA data

Step 7: Get Variants & Disease

• gnomad_get_gene_constraints → Genetic constraint scores
• gnomad_get_gene → Population variants
• ClinVar_search_variants → Clinical variants
• UniProt_get_disease_variants_by_accession → Disease variants
• OpenTargets_get_diseases_phenotypes_by_target_ensembl → Disease associations

Step 8: Get Drug Interactions

• OpenTargets_get_target_tractability_by_ensemblID → Druggability assessment
• DGIdb_get_gene_druggability → DGIdb druggability
• OpenTargets_get_associated_drugs_by_target_ensemblID → Known drugs
• ChEMBL_get_target_activities → Bioactivity data
• OpenTargets_get_target_safety_profile_by_ensemblID → Safety liabilities

Step 9: Get Literature

• PubMed_search_articles → PubMed publications
• OpenTargets_get_publications_by_target_ensemblID → Target-specific papers

---

Use Case 2: Drug/Compound Research

Step 1: Resolve Compound Identity

• PubChem_get_CID_by_compound_name → Get PubChem CID
• ChEMBL_search_molecules → Get ChEMBL ID
• DailyMed_search_spls → Check if approved drug
• PharmGKB_search_drugs → Get PharmGKB ID

Step 2: Get Chemical Properties

• PubChem_get_compound_properties_by_CID → Molecular properties
• ADMETAI_predict_physicochemical_properties → Predicted properties
• ADMETAI_predict_solubility_lipophilicity_hydration → Solubility data

Step 3: Get Targets & Bioactivity

• ChEMBL_search_activities → Bioactivity data
• ChEMBL_get_target → Target proteins
• DGIdb_get_drug_info → Drug-gene interactions
• PubChem_get_compound_bioactivity → PubChem bioactivity

Step 4: Get ADMET Predictions

• ADMETAI_predict_bioavailability → Absorption predictions
• ADMETAI_predict_BBB_penetrance → BBB penetration
• ADMETAI_predict_CYP_interactions → CYP metabolism
• ADMETAI_predict_clearance_distribution → Distribution/clearance
• ADMETAI_predict_toxicity → Toxicity predictions

Step 5: Get Clinical Trials

• search_clinical_trials → Find clinical trials
• get_clinical_trial_conditions_and_interventions → Trial details
• extract_clinical_trial_outcomes → Trial outcomes
• extract_clinical_trial_adverse_events → Safety data

Step 6: Get Safety Data

• FAERS_count_reactions_by_drug_event → Adverse events
• FAERS_count_seriousness_by_drug_event → Seriousness distribution
• FAERS_count_outcomes_by_drug_event → Outcome distribution
• OpenTargets_get_drug_warnings_by_chemblId → Drug warnings

Step 7: Get Pharmacogenomics

• PharmGKB_get_drug_details → PharmGKB drug info
• PharmGKB_get_clinical_annotations → Clinical annotations
• PharmGKB_get_dosing_guidelines → Dosing guidelines

---

Use Case 3: Disease Research

Step 1: Resolve Disease Identity

• OSL_get_efo_id_by_disease_name → Get EFO ID
• OpenTargets_get_disease_id_description_by_name → Disease info
• ols_search_efo_terms → Search EFO ontology
• umls_search_concepts → UMLS concept
• icd_search_codes → ICD-10 code
• snomed_search_concepts → SNOMED CT

Step 2: Get Phenotypes

• OpenTargets_get_associated_phenotypes_by_disease_efoId → Phenotypes
• get_HPO_ID_by_phenotype → HPO terms
• MedlinePlus_search_topics_by_keyword → Patient info
• MedlinePlus_get_genetics_condition_by_name → Genetics info

Step 3: Get Associated Genes

• OpenTargets_get_associated_targets_by_disease_efoId → Associated genes
• OpenTargets_target_disease_evidence → Evidence details
• ClinVar_search_variants → Pathogenic variants

Step 4: Get GWAS Associations

• gwas_search_associations → GWAS hits
• gwas_get_variants_for_trait → Associated variants
• gwas_get_associations_for_trait → Association details
• gwas_get_studies_for_trait → GWAS studies

Step 5: Get Treatment Options

• OpenTargets_get_associated_drugs_by_disease_efoId → Approved/trial drugs
• search_clinical_trials → Clinical trials
• GtoPdb_search_diseases → Guide to Pharmacology

Step 6: Get Pathways

• Reactome_get_diseases → Disease pathways
• Reactome_map_uniprot_to_pathways → Protein pathways
• humanbase_ppi_analysis → Tissue-specific networks
• geo_search_datasets → Expression datasets

Step 7: Get Literature

• PubMed_search_articles → Publications
• OpenTargets_get_publications_by_disease_efoId → Disease papers
• openalex_search_works → OpenAlex literature

Step 8: Get Similar Diseases

• OpenTargets_get_similar_entities_by_disease_efoId → Similar diseases

---

Complete Fallback Chains

Literature Tools

Protein/Gene Tools

Drug/Compound Tools

Disease Tools

Clinical Tools

---

ID Cross-Reference Matrix

Gene/Protein ID Conversions

Compound ID Conversions

Disease ID Conversions

---

Example Research Workflows

Example 1: "Tell me about EGFR"

1. Disambiguate: Find UniProt ID (P00533), Ensembl ID (ENSG00000146648) 2. Parallel queries:

• Get UniProt entry
• Get domain architecture (InterPro)
• Get AlphaFold structure
• Get protein interactions (STRING)
• Get tissue expression (GTEx)
• Get genetic constraints (gnomAD)
• Get druggability (OpenTargets)
• Get known drugs (OpenTargets)

3. Deep dive: Get PDB structures for drug-bound forms 4. Literature: Search PubMed for "EGFR AND cancer" 5. Synthesize: Create comprehensive report with all findings

Example 2: "What drugs target TP53?"

1. Get target info: Ensembl ID ENSG00000141510 2. Get drug associations: OpenTargets drugs, DGIdb interactions 3. For each drug:

• Get ChEMBL compound details
• Get bioactivity data
• Get clinical trials

4. Check tractability: Is TP53 druggable? 5. Synthesize: Report on drug landscape with evidence

Example 3: "Research Alzheimer's disease"

1. Disambiguate: EFO_0000249 2. Get ontology: Disease subtypes, synonyms 3. Get targets: Top 50 associated genes 4. For top targets: Get evidence details 5. Get phenotypes: HPO terms 6. Get drugs: Approved and trial drugs 7. Get clinical trials: Current studies 8. Get variants: ClinVar pathogenic variants 9. Get GWAS: Associated loci 10. Get literature: Recent publications 11. Synthesize: Comprehensive disease report

---

Database Coverage Summary

---

Report Section Templates

Target Report Sections

1. Executive Summary 2. Target Identifiers 3. Basic Information (name, function, localization) 4. Structural Biology (PDB, AlphaFold, domains) 5. Function & Pathways (GO, Reactome, KEGG) 6. Protein-Protein Interactions 7. Expression Profile 8. Genetic Variation & Disease 9. Druggability & Pharmacology 10. Safety Profile 11. Literature & Research Landscape 12. Competitive Landscape 13. Summary & Recommendations 14. Data Sources & Methodology

Drug Report Sections

1. Executive Summary 2. Compound Identity 3. Chemical Properties 4. Mechanism & Targets 5. ADMET Properties 6. Clinical Development 7. Safety Profile 8. Pharmacogenomics 9. Regulatory & Labeling 10. Literature & Research 11. Conclusions 12. Data Sources

Disease Report Sections

1. Executive Summary 2. Disease Identity & Classification 3. Clinical Presentation 4. Genetic & Molecular Basis 5. Treatment Landscape 6. Biological Pathways 7. Epidemiology & Risk Factors 8. Literature & Research Activity 9. Similar Diseases & Comorbidities 10. Data Sources

General Usage Strategies

Use these strategies ONLY when no specialized skill matches the user's query. See SKILL.md routing table first.

Core Principles

1. Search widely — run multiple tool discovery queries with synonyms 2. Query multiple databases — cross-reference across sources 3. Multi-hop persistence — chain 5-10 tool calls; one is rarely enough 4. Never give up — if a tool fails, try alternatives 5. Comprehensive reports — cite sources, grade evidence 6. English-first queries — translate to English for tools, respond in user's language

---

Strategy 0: Clarify Before Acting

Ask clarifying questions when:

• Vague entity ("Research cancer" → which type?)
• Ambiguous name ("JAK" → JAK1/2/3? Inhibitor?)
• Unclear scope ("Look into metformin" → safety? repurposing? profile?)
• Multiple interpretations ("ACE" → gene? inhibitors? ACE2?)

Do NOT ask when specific enough: "What is the structure of EGFR kinase domain?"

Checklist: confirm entity, species (default human), scope, output format.

---

Strategy 1: Exhaustive Tool Discovery

Run at least 6 queries in parallel: 1. Main topic: find_tools(query="[topic]") 2. Synonym: find_tools(query="[synonym]") 3. Another synonym: find_tools(query="[synonym2]") 4. Known database: find_tools(query="[database name]") 5. Data type: find_tools(query="[data type]") 6. Full use case: find_tools(query="[full description]")

Actually RUN these, don't just describe them.

---

Strategy 2: Multi-Hop Tool Chains

Common patterns:

Rules: Don't stop at first result. Follow cross-references. 5-10 calls is normal. Track all IDs.

---

Strategy 3: Query Multiple Databases

Merge: collect all, note sources, handle conflicts, prefer curated data.

---

Strategy 3.1: Full-Text Literature Search

For body-only terms (rsIDs, figure refs, supplementary tables) that don't appear in abstracts:

1. PMC_search_papers — NCBI PMC indexes full text 2. EuropePMC_search_articles with require_has_ft=true + fulltext_terms=[...] 3. EuropePMC_get_fulltext_snippets — confirm term is in paper 4. CORE_get_fulltext_snippets — PDF scan fallback

---

Strategy 4: Disambiguation

Before research, establish canonical IDs:

• Genes: Symbol → UniProt, Ensembl, NCBI Gene, ChEMBL target
• Compounds: Name → PubChem CID, ChEMBL ID, SMILES
• Diseases: Name → EFO ID, ICD-10, UMLS CUI

Gather synonyms. Detect naming collisions. Confirm species (default: human).

---

Strategy 5: Failure Handling

Primary tool → fails → retry → fails → fallback #1 → fails → fallback #2 → document as unavailable

Common fallbacks:

• PubMed → EuropePMC → OpenAlex
• GTEx → Human Protein Atlas
• PubChem → ChEMBL → SMILES-based
• UniProt → Proteins API

Also try: synonyms, broader terms, different databases.

---

Strategy 6: Comprehensive Reports

1. Create report structure first, then fill progressively 2. Every fact needs a source citation 3. Grade evidence: ★★★ (mechanistic) → ★★☆ (functional) → ★☆☆ (association) → ☆☆☆ (review) 4. All sections must exist, even if "data unavailable" 5. Quality target: 15-30+ tool calls, all sections filled

---

Strategy 7: Defer to Specialized Skills

If you realize mid-research that a specialized skill matches: STOP general strategies, ROUTE to the skill.

---

Strategy 8: Parallel Execution

Run independent queries simultaneously:

• Different databases for same entity → parallel
• Tool B needs output from Tool A → sequential

---

Strategy 9: Completeness Check

After gathering data, ask: "What's still missing?"

• All identifiers? Core data? Context? Relationships? Variants? Evidence? Literature? Gaps documented?
• Stop when: all aspects addressed, multiple sources queried, gaps documented.

---

Strategy 10: English-First Queries

All tool calls use English. Translate non-English input before querying. Respond in user's language.

---

Troubleshooting