01Safety
0.0 / 30
What changed in the harness
Selection accuracy 98→100, token cost up 8%, 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
27.8 / 30
03Economics
20.0 / 20
04Discoverability
10.0 / 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 |
|---|---|---|
comet_connect |
name_restates_behavior no_return_description |
Establish a live connection to the Comet browser, auto-starting it if it is not already running and blocking until it is ready. Returns the connection status once the browser is available. |
comet_poll |
no_return_description |
Check the status and progress of the currently running agentic task; call repeatedly to monitor it until completion. Returns the latest status, progress, and any partial results. |
comet_stop |
no_return_description |
Halt the currently running agent task, useful when it is going off track or no longer needed. Returns confirmation that the task was stopped. |
comet_screenshot |
no_return_description |
Capture a screenshot of the page currently displayed in the browser. Returns the screenshot as image data. |
comet_mode |
no_return_description |
Switch Perplexity search mode between 'search' (basic), 'research' (deep research), 'labs' (analytics/visualization), and 'learn' (educational), or omit the mode parameter to view the current mode. Returns the active mode after the operation. |
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 |
|---|---|---|---|
comet_ask |
comet_mode |
medium | Both definitions mention 'deep research': comet_ask touts 'deep research with agentic browsing,' while comet_mode's 'research' mode is described as 'deep research.' A natural-language task like 'do deep research on X in Comet' could cause the agent to either call comet_mode to switch to research mode or call comet_ask directly, since ask already handles deep research. The shared perplexity token reinforces the overlap. |
Compare the field