feat(harness): null + random-click baselines — the benchmark floor (closes #245)#257
Open
reacher-z wants to merge 1 commit into
Open
feat(harness): null + random-click baselines — the benchmark floor (closes #245)#257reacher-z wants to merge 1 commit into
reacher-z wants to merge 1 commit into
Conversation
closes #245) Two leaderboard baselines that establish the ClawBench floor and validate the interceptor: - null: takes no action → 0 actions/requests, interceptor never fires. A do-nothing agent must score ~0 (proves the benchmark isn't luck-passable and the HTTP interceptor has no false positives without genuine activity). - random-click: dispatches N random CDP mouse clicks (websocket-client) with no task intent → a non-trivial floor; browsing activity alone, without hitting the target action, must not pass. Reads the shared CLAWBENCH_BROWSER_CDP_URL. Both registered in harnesses.yaml (null quoted to dodge the YAML null keyword) + test_harness_registry. The interceptor false-positive rate over these baselines is reported by clawbench-analyze (#159). Verified on real hardware: registry test passes; both images build from clawbench-base; null run-harness is a no-op; the random-clicker connects to Chromium via CDP and dispatches+logs 10 clicks. Full suite green; ruff clean.
Perry2004
approved these changes
Jul 11, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Null + random-click baselines — the benchmark floor (closes #245)
Two leaderboard baselines that establish the ClawBench floor and validate the interceptor (the methodological artifacts reviewers ask for):
null— takes no action → 0 actions/requests, the interceptor never fires. A do-nothing agent must score ~0, proving ClawBench isn't luck-passable and that Stage-1 has no false positives without genuine activity.random-click— dispatches N random CDP mouse clicks (viawebsocket-client) with no task intent → a non-trivial floor: browsing activity alone, without hitting the target action, must not pass. Reads the sharedCLAWBENCH_BROWSER_CDP_URL.Both registered in
harnesses.yaml(+test_harness_registry); the interceptor false-positive rate over these baselines is computed byclawbench-analyze(#159).Verified on real hardware (podman): registry test passes; both images build from
clawbench-base;null's run-harness is a genuine no-op; the random-clicker connects to Chromium via CDP and dispatches + logs 10 clicks. Full suite green; ruff clean.Supersedes #256 (null-only) — this folds in the random-click baseline too.