1.0.129MIT
strict
core24
run ai coding agents on eval tasks in sandboxes, and score them
pats runs ai coding agents (the claude, codex, copilot and opencode clis) on
eval tasks, each in a docker or lxd sandbox behind an egress proxy, and
scores what they produce.
this snap is the strict build: docker and local lxd only, through the
docker and lxc clients it ships, and every step in a step sandbox. connect
the engine you use:
eval tasks, each in a docker or lxd sandbox behind an egress proxy, and
scores what they produce.
this snap is the strict build: docker and local lxd only, through the
docker and lxc clients it ships, and every step in a step sandbox. connect
the engine you use:
snap connect pats:docker (or
pats:docker docker:docker-daemon for the docker snap), or
snap connect pats:lxd lxd:lxd.Update History
1.0.0 (22) → 1.0.1 (29)14 Nov 2026, 10:00 UTC
1.0.0-beta.15 (20) → 1.0.0 (22)13 Nov 2026, 10:30 UTC
1.0.0-beta.14 (10) → 1.0.0-beta.15 (20)13 Nov 2026, 10:15 UTC
1.0.0-beta.13 (6) → 1.0.0-beta.14 (10)13 Nov 2026, 08:15 UTC
1.0.0-beta.12+git2a61170 (2) → 1.0.0-beta.13 (6)12 Nov 2026, 00:15 UTC
1.0.0-beta.12+git2a61170 (2)11 Nov 2026, 23:30 UTC
11 Nov 2026, 01:05 UTC
13 Nov 2026, 17:05 UTC
11 Nov 2026, 23:30 UTC