01 / PROJECTS

Work, with the details.

Two projects around how agents run, how we evaluate them, and what their results actually tell us.

01 / TOOLING · EVALUATIONGitHub

skillfit

Compare Skills, rules, and MCP setups on the same tasks. Keep quality, trigger behavior, and token cost in view.

SAME TASKS · PAIRED RUNS
ABaseline
/
BTreatment
QualityBehaviorToken cost
02 / HARNESS · TOOL USEGitHub

DeviceAgent Arena

A testbed for Chinese tool-use agents, where actions change the next state. Check both the outcome and the path taken.

STATEFUL TASKS · CLOSED LOOP
TaskHarnessTools
observation → next decision
Final stateFull trajectory

02 / OPEN SOURCE

Small fixes, shared upstream.

Selected contributions to tools I use and care about.

Contribution history