01Safety
0.0 / 30
What changed in the harness
Selection accuracy 100→98, token cost up 2%, 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
29.2 / 30
03Economics
20.0 / 20
04Discoverability
12.5 / 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.
1 defect 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 |
|---|---|---|
billing_status |
name_restates_behavior no_return_description |
Returns your current plan name and remaining credit balance. Call it for free with no parameters to verify you have enough credits before running a paid search. |
Selection evidence
Confusable tool pairs.
1 pair where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
search |
get_search |
medium | A task phrased as 'get the search results' or 'retrieve a search' could lexically route an agent to get_search, but that tool requires a search_id from a prior search call and only re-reads existing searches, whereas search actually runs a new query. If the user says 'get me a search of X', an agent might plausibly pick get_search on name similarity alone, though the descriptions disambiguate once the search_id requirement is noticed. |
Compare the field