NOP Agent (No Operation)
The NOP Agent (No Operation) is a baseline agent that performs no work. It serves as the lower-bound performance baseline for Harbor evaluations.
The NOP agent is used for benchmarking to validate that an evaluation pipeline is wired correctly and that failures are not caused by agent execution itself.
-
No setup Does not install dependencies or perform any environment initialization.
-
No execution Does not process the instruction or execute any task logic.
-
No results Returns immediately with an empty context after running the
test.shfile. -
0% success rate (expected) By definition, the NOP agent solves nothing. All task evaluations are expected to fail, and this failure is considered a pass for the CI step. If the NOP agent passes a task, the task definition or evaluation logic is likely incorrect.
-
CI behavior If the NOP agent completes evaluation successfully (i.e., fails all tasks as expected), subsequent agents can run without requiring execution of the user’s
solve.shfile.