01Safety
0.0 / 30
What changed in the harness
Selection accuracy 100→100, token cost up 4%, unconfirmed writes 0%→0%.
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
27.2 / 30
03Economics
19.8 / 20
04Discoverability
12.9 / 20
Highest-impact fix
Estimated gain +30 pointsAdd explicit identity and permission preflight tools
Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.
Description evidence
Defects and rewrites.
5 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 |
|---|---|---|
create_gift_card |
no_return_description |
Create a gift card by debiting the customer's DAH balance. Requires dual signatures (customer + distribution). The card value is escrowed until the recipient claims it. Returns the created card details, including its unique identifier and ISSUED status. |
claim_gift_card |
no_return_description |
Claim a gift card and credit the claimer's DAH balance. Only cards with status ISSUED can be claimed. Returns the claimed card details and the updated DAH balance. |
create_sponsor_card |
no_return_description |
Create a location-based sponsor card by debiting the sponsor's DAH balance. Requires dual signatures. The card can be claimed by users near the sponsor location. Returns the created sponsor card details, including its unique identifier and ISSUED status. |
claim_sponsor_card |
no_return_description |
Claim a sponsor card and credit the claimer's DAH balance. Only cards with status ISSUED can be claimed. Returns the claimed sponsor card details and the updated DAH balance. |
purchase_product |
name_restates_behavior no_return_description |
Debit the sender's DAH balance and credit the receiver (merchant) to complete a product purchase. Requires dual signatures (buyer + merchant). Returns the transaction confirmation and order details. |
Selection evidence
Confusable tool pairs.
4 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
claim_gift_card |
claim_sponsor_card |
high | Both definitions are near-identical (same action, same claimer credit, same required fields), differing only in card type. A task like 'claim my card' or 'redeem this card' without naming gift vs sponsor leaves the type ambiguous, so the agent could pick the wrong claim tool. |
create_gift_card |
claim_gift_card |
medium | Same gift-card entity with opposite actions: create debits the giver with dual signatures, claim credits the receiver. A generically phrased task like 'handle this gift card' or 'process the gift card details' does not reveal whether the user is funding or redeeming it, making mis-selection plausible. |
create_gift_card |
create_sponsor_card |
medium | Both create a card by debiting balance via dual signatures (shared tokens 'create', 'funding', 'debiting'). A task like 'create a card' or 'issue my card' does not state whether it is customer-funded (gift) or sponsor-funded and location-based (sponsor), so the agent may pick the wrong creator. |
create_sponsor_card |
claim_sponsor_card |
medium | Same sponsor-card entity with opposite actions and identical required claimer fields. A task phrased without an explicit action verb, e.g. 'handle this sponsor card' or 'sponsor card request', could plausibly be read as either creating or claiming it. |
Compare the field