agent-bench: framework benchmark rig + aider & mini-SWE-agent comparands - #2

Open

PositiveControl wants to merge 11 commits into

mainfrom

worktree-agent-framework-bench

Commits on Jun 15, 2026

docs: agent-framework benchmark plan (GTA build target, gx10 vLLM)

PositiveControl
and
claude
committed
7c6ffe3
agent-bench: scaffold runner + aider adapter + three scorers

PositiveControl
and
claude
committed
98f83e1
agent-bench: opencode + goose adapters, cost scorer, shared _proc

PositiveControl
and
claude
committed
4c45239
agent-bench: exact token extraction from opencode --format json

PositiveControl
and
claude
committed
6d5c55a
agent-bench: process scorer (diff LOC, turns, tool calls, tracebacks)

PositiveControl
and
claude
committed
2d374a7
agent-bench: pivot to HTML5-canvas web game + live gx10 validation

PositiveControl
and
claude
committed
1393a8c
agent-bench: resolve html entry to absolute before as_uri()

PositiveControl
and
claude
committed
8ee2193
agent-bench: Bar-A tracer-bullet close (gx10 FP8 pin + clean status)

PositiveControl
and
claude
committed
b53cbc5
agent-bench: mini-SWE-agent adapter (headless via Python API)

PositiveControl
and
claude
committed
9bcd2f5
agent-bench: bring rig up to repo standards (mypy + tests)

PositiveControl
and
claude
committed
c755f00
test: isolate bd_warning_filter tests from inherited GIT_* env

PositiveControl
and
claude
committed
b8d721a