01Safety
0.0 / 30
What changed in the harness
Selection accuracy 69→60, token cost up 1%, 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
22.9 / 30
03Economics
16.6 / 20
04Discoverability
13.0 / 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.
4 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 |
|---|---|---|
gpu_run |
no_return_description |
Submit a job to run the specified GPU-Bridge AI service (LLM, image/video generation, speech-to-text, TTS, embeddings, OCR, and other services) with the given service-specific input, and return a job identifier along with status information; use gpu_status to poll for and retrieve the final results, and gpu_catalog to see all available services. |
gpu_catalog |
no_return_description |
List all available GPU-Bridge services and return, for each, its service key, pricing, and model info. No authentication required. |
gpu_balance |
no_return_description |
Check the account's GPU-Bridge status and return current credit balance, today's spend, the applicable volume discount tier, and recent job history. |
gpu_estimate |
name_restates_behavior no_return_description |
Compute a projected cost for running a given GPU-Bridge service before submitting it via gpu_run, optionally based on an estimated runtime in seconds, and return the estimated price. No authentication required. |
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 |
|---|---|---|---|
gpu_catalog |
gpu_estimate |
low | Both relate to pricing info, but catalog lists all services with pricing while estimate computes cost for one specific service; a task like 'how much would this cost' could plausibly trigger either. |
gpu_status |
gpu_balance |
low | Both are 'check' style informational queries, but status is job-specific (requires job_id) and balance is account-wide; distinct enough that confusion is unlikely except for vague 'check my GPU-Bridge account/job' phrasing. |
Compare the field