Cold One-Shot Capability Benchmark
Can an external agent given a genuinely one-off task conclude that Hydracept is useful because it gets from the request to the result with negligible extra coordination?
The agent is not told to use Hydracept. Repeated usage must not be required to justify the first call.
Make me a transparent 1024x1024 icon of a blue slime. Use whatever capability makes sense. This may be the only generated asset I ever need.
A second gate: recreating projects or workspaces must not reset managed first-use
entitlement. Exhaustion returns funding_required, not a generic failure.
BYOK and funded managed execution remain available.
Read the methodology
or run python scripts/run_cold_one_shot_benchmark.py for the structural gate.