Loading ToolBench
Food additive safety regulatory reference data: verified safety scores, ADI values, JECFA/EFSA evaluations, allergen and dietary flags, EU/US/Israel regulatory status, Israeli nutrition profiles, and pesticide MRL limits. 6,450+ additives, 6,563 JECFA evaluations, 5,251 EFSA substances, 77,278 synonyms, 4,624 nutrition profiles, 3,708 pesticide MRLs.
Five tools with strong descriptions (194 - 350 chars, well above baseline 194 avg) and clear use-case guidance. All tools have input schemas with types and descriptions. Tool names follow verb_noun pattern (check_*, search_*). Enum constraints present for market and filter parameters. However, output schemas are NOT documented in the source, descriptions mention return fields but no formal schema definitions are visible. Parameter descriptions are good but lack explicit format/range constraints (e.g., 'up to 60 ingredients' is stated but not enforced in schema). No error handling guidance visible. Tool annotations (readOnlyHint) are correctly applied. Rate limiting and auth logic are implemented server-side, which is correct.
Retrieve verified food additive safety data including toxicological evaluations, regulatory status, and dietary compatibility from 6,450+ food additive records. Input: additive common name, E-number, or CAS number (e.g. 'aspartame', 'E951', '22839-47-0'). Returns: safety score (1-10 scale), ADI (Acceptable Daily Intake) with source, JECFA evaluation history (ADI status, functional class, evaluation year), EFSA toxicology data (ADI/TDI/NOAEL values, genotoxicity), EU and US regulatory status, Israel MOH permitted-additive status, IARC classification, health concerns, allergen flags, dietary compatibility (vegan, halal, kosher, diabetic), PubChem chemistry data, common foods, and banned-country list. Sources: JECFA (6,563 evaluations), EFSA (5,251 substances), EU food additive regulations, US FDA GRAS, Israel MOH, IARC. For food additive safety assessment and regulatory compliance. Do not use for full ingredient-list scans (use check_ingredient_list), nutrition queries (use check_nutrition), or pesticide MRLs (use check_pesticide_mrl).
Scan a packaged-food product ingredient list for additive safety risks, regulatory compliance, allergens, and dietary compatibility across target markets. Input: comma-separated ingredient list as printed on food packaging (up to 60 ingredients). Returns per-ingredient: matched additive name, E-number, safety score (1-10), category, EU/US status, health concerns, allergen flags, and ADI. Returns overall: LOW/MODERATE/HIGH risk assessment, average safety score, flagged high-risk additives with reasons and banned-country data, allergen warnings, and market-specific compliance notes. Database: 6,450+ food additives with 77,278 synonyms for fuzzy matching. For food product safety screening and market compliance. Do not use for single additive lookup (use check_additive), nutrition facts (use check_nutrition), or pesticide MRLs (use check_pesticide_mrl).
Output schemas not documented. Descriptions mention return fields (safety score, ADI, JECFA evaluations, etc.) but no formal JSON Schema definitions are visible in source. LLMs cannot plan downstream tool calls or extract structured data without knowing the response shape.
Parameter constraints not formalized in schema. Descriptions state 'up to 60 ingredients', 'up to 120 chars', 'up to 25 results' but maxLength/maxItems are not visible in the input schema definitions. Validation happens server-side (normalizeQuery, MAX_INGREDIENTS constants) but schema does not enforce these.
No error handling guidance in tool descriptions. Descriptions do not explain what happens on invalid input (e.g., 'additive not found', 'invalid market code'), what the LLM should do next, or whether errors are retryable. Agents cannot self-correct without explicit recovery hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
Retrieve Israeli Ministry of Health official nutrition data for a food item by name in Hebrew or English. Input: food name (e.g. 'hummus', 'chicken breast', 'bread', or Hebrew equivalent). Returns per 100g: energy (kcal), protein, total fat, carbohydrates, dietary fiber, total sugars, alcohol, moisture; vitamins (A, C, E, D, K, B6, B12, thiamin, riboflavin, niacin, folate); minerals (calcium, iron, magnesium, phosphorus, zinc, selenium, sodium, potassium, choline); fats (cholesterol, saturated, mono/polyunsaturated, trans). Source: Israel Ministry of Health Nutrition Database (4,624 food items). For nutritional analysis and dietary reference. Do not use for additive safety (use check_additive) or pesticide MRLs (use check_pesticide_mrl).
Check Israeli official pesticide Maximum Residue Limits (MRL) by pesticide active substance, crop, or pesticide-crop pair. Input: pesticide name, crop name, or combined query in Hebrew or English (e.g. 'glyphosate', 'tomato', 'chlorpyrifos apple'). Returns: active substance, crop (Hebrew and English), official MRL value in mg/kg (ppm), last update date, and pending regulatory changes. Source: Israel Ministry of Health Pesticide Residue Limits (3,708 pesticide-crop MRL records). For agricultural compliance and food safety residue assessment. Do not use for food additive safety (use check_additive), cannabis pesticide limits, or non-Israeli MRL regimes.
Search food additive records by keyword, function, category, dietary filter, or health concern for additive discovery and selection. Input: food additive keyword (e.g. 'preservative', 'sweetener', 'hyperactivity', 'banned', 'E1') with optional filter (high_risk, allergens, banned, not_vegan, not_halal). Returns: matching additive names, E-numbers, CAS numbers, Hebrew names, categories, functions, safety scores (1-10), EU/US status, health concerns, allergen flags, dietary status (vegan/halal/kosher), and ADI values. Database: 6,450+ food additives searchable by name, function, category, and Hebrew name. For food additive discovery and category exploration. Do not use when the user has an exact additive name/E-number for full safety data (use check_additive).
No pagination guidance for list-returning tools. search_additives and check_ingredient_list return multiple results (up to 25 and per-ingredient respectively) but no limit/offset or cursor parameters are documented. Large result sets risk context window exhaustion.
Tool descriptions include example values (e.g., 'aspartame', 'E951', '22839-47-0', 'hummus', 'glyphosate'). LLMs tend to reuse examples literally rather than adapting to context, risking incorrect calls with hardcoded example data.