01Safety
30.0 / 30
What changed in the harness
Selection accuracy 95→98, token cost up 6%, 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
28.0 / 30
03Economics
20.0 / 20
04Discoverability
12.1 / 20
Highest-impact fix
Estimated gain +8 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.
3 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 |
|---|---|---|
highlight_element |
no_return_description |
Briefly flash-highlight the element with the given name on the annotated page so the user can visually confirm which element you are referring to. Returns a confirmation that the highlight was triggered, or an error if the element name is not found. |
rescan_elements |
no_return_description |
Force the annotated page to rescan all UI elements. Use this after the page content has changed (e.g. after navigation, DOM updates, or user interaction) so the detected element set stays current. Returns the updated list of detected elements, including names, CSS selectors, positions, and sizes. |
inspect_mode |
no_return_description |
Toggle inspect mode on the annotated page. When ON, the user can click any element to copy its name; when OFF, the page behaves normally. Use this to help the user copy element names for communicating design changes. Returns the new inspect mode state (true if enabled, false if disabled). |
Selection evidence
Confusable tool pairs.
2 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
get_elements |
rescan_elements |
medium | Tasks like "refresh the page elements" or "get the latest elements after the page changed" are ambiguous: the agent may call get_elements to fetch current annotations, or rescan_elements to re-detect. The boundary between retrieving elements and forcing a re-detection is not clear from descriptions alone. |
get_elements |
highlight_element |
low | Phrasing like "show me the element" or "point out the element on the page" could be read either as listing what elements exist (get_elements, which explains what is on the page) or visually flashing one (highlight_element). The mismatch is plausible but weak since highlight requires a specific name while get_elements returns the full set. |
Compare the field