01Safety
30.0 / 30
What changed in the harness
Selection accuracy 100→98, token cost up 7%, unconfirmed writes 100%→100%.
Category breakdown
Where the score comes from.
Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.
02Legibility
27.5 / 30
03Economics
18.6 / 20
04Discoverability
15.4 / 20
Highest-impact fix
Estimated gain +5 pointsMake target tools discoverable on the first call
Clarify tool names, decision boundaries, and required argument schemas so an agent can choose and construct the target call without exploratory steps.
Description evidence
Defects and rewrites.
2 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.
| Tool | Defect types | Suggested rewrite |
|---|---|---|
calculate_us_federal_tax |
params_unexplained |
Calculate US federal income tax for tax year 2026 using the IRS progressive brackets and the standard deduction for the chosen filing status. Use for a quick federal-only estimate; it does NOT include state income tax or FICA payroll taxes. Inputs: `income` (annual gross income in USD) and optional `filing_status` — the IRS filing status, which determines the applicable standard deduction and rate brackets: 'single' (default), 'married_joint' (married filing jointly), 'married_separate' (married filing separately), or 'head_of_household'. Returns taxable income after standard deduction, federal tax, effective and marginal rates. Read-only and deterministic. Estimates only — not tax advice. |
calculate_tdee |
params_unexplained |
Calculate Total Daily Energy Expenditure (maintenance calories) by multiplying a Basal Metabolic Rate by a standard activity factor. Inputs: `bmr` (basal metabolic rate in kcal — compute it first if unknown) and `activity_level`, the daily-activity multiplier applied to BMR: 'sedentary' (little or no exercise), 'light' (light exercise 1–3 days/week), 'moderate' (moderate exercise 3–5 days/week), 'active' (hard exercise 6–7 days/week), or 'very_active' (intense daily training or a physical job). Returns TDEE in kcal/day. Read-only and deterministic. General guidance, not medical or dietary advice. |
Selection evidence
Confusable tool pairs.
9 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
calculate_mortgage |
calculate_loan_payment |
high | Both compute the fixed monthly payment of an amortizing loan with near-identical inputs and output; a task like 'what is my monthly payment on this loan/mortgage' that omits whether the term is in months or years (or needs a schedule) can land on either tool. |
calculate_us_federal_tax |
calculate_canada_federal_tax |
medium | Both are named 'federal' tax calculators with progressive brackets; a natural task like 'what is my federal income tax' without a country is genuinely ambiguous between the two. |
calculate_french_income_tax |
calculate_uk_income_tax |
medium | Both compute personal income tax from annual income using progressive brackets; a country-less task like 'how much income tax do I owe for the year' could select either, and both share the 'personal/tax' terminology. |
calculate_french_income_tax |
calculate_french_salary |
medium | Both concern French salary/tax outcomes; a task about 'what do I take home from my French gross salary' or 'what taxes are on my salary' could trigger either, since salary does gross-to-net and tax computes impôt, despite the cross-references. |
list_bundles |
get_bundle_tools |
medium | Users phrase discovery tasks loosely ('what calculators do you have', 'find me a tool for X'); list_bundles only lists thematic bundles while get_bundle_tools lists the tools inside one, and either could be called first. |
calculate_french_income_tax |
calculate_us_federal_tax |
low | Only an underspecified task like 'calculate my income tax for 2026' with no country leaves both candidates; otherwise the France/US distinction is explicit in names and descriptions. |
calculate_french_income_tax |
calculate_canada_federal_tax |
low | Confusion requires a country-less 'income tax' request; tool names and the EUR/CAD input parameters otherwise clearly separate France and Canada. |
calculate_uk_income_tax |
calculate_canada_federal_tax |
low | Both are income-tax calculators, but GBP vs CAD inputs and country keywords disambiguate unless a loose task omits the country entirely. |
calculate_us_federal_tax |
calculate_uk_income_tax |
low | Same pattern: a generic 'income tax' task without a country could match either, but the US vs UK distinctions in names, currency, and filing-status inputs otherwise resolve the choice. |
Compare the field