Kyvero-Vexus/agent-ab-testing
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Activity
OpenClaw skill for fair A/B testing of agent conditions
2 commits 1 branch 0 tags 45 KiB
Find a file
Exact
chrysolambda-ops 90aa5d4d98 docs: add AGENTS contribution and PR workflow guide
2026-04-02 00:09:52 -04:00
notes feat: add agent A/B testing OpenClaw skill 2026-03-08 13:34:25 -04:00
templates feat: add agent A/B testing OpenClaw skill 2026-03-08 13:34:25 -04:00
AGENTS.md docs: add AGENTS contribution and PR workflow guide 2026-04-02 00:09:52 -04:00
LICENSE feat: add agent A/B testing OpenClaw skill 2026-03-08 13:34:25 -04:00
README.md feat: add agent A/B testing OpenClaw skill 2026-03-08 13:34:25 -04:00
SKILL.md feat: add agent A/B testing OpenClaw skill 2026-03-08 13:34:25 -04:00

README.md

agent-ab-testing

OpenClaw skill for designing and running fair A/B tests between two agent conditions.

What it covers

  • defining the variable under test
  • controlling invariants
  • benchmark design
  • harness validation
  • serial subagent execution
  • objective grading
  • blind rubric judging
  • pitfalls like ceiling effects, leakage, and harness asymmetry
  • iteration and lab-note discipline

Main file

  • SKILL.md

Companion materials

  • templates/ab-test-report-template.md
  • templates/judge-packet-template.md
  • notes/lessons-learned.md

License

AGPL-3.0-or-later