OpenSpec proposal: openspec/changes/add-counterfactual-attribution/
Single big feature
A4 Counterfactual causal LCP attribution — turns Lighthouse-style 'potential savings' estimates into MEASURED savings. After baseline N runs, pick top-3 candidates, run intervention pass via CDP Fetch.failRequest (remove) / Fetch.fulfillRequest (preload-simulate) / Network.setBlockedURLs (block 3P). Mann-Whitney U significance per candidate. 4-verdict output: causal-with-high-confidence | causal-with-medium-confidence | inconclusive | no-effect.
Status
Phase 5 OpenSpec artifacts COMMITTED (a16c3c5). 1082 LOC. 14 requirements / 30 scenarios. All 5 design questions resolved.
Effort
L (~2 weeks)
Dependencies
🤖 Generated by harness-init dogfood + Sisyphus v2 strategy
OpenSpec proposal:
openspec/changes/add-counterfactual-attribution/Single big feature
A4 Counterfactual causal LCP attribution — turns Lighthouse-style 'potential savings' estimates into MEASURED savings. After baseline N runs, pick top-3 candidates, run intervention pass via CDP
Fetch.failRequest(remove) /Fetch.fulfillRequest(preload-simulate) /Network.setBlockedURLs(block 3P). Mann-Whitney U significance per candidate. 4-verdict output:causal-with-high-confidence | causal-with-medium-confidence | inconclusive | no-effect.Status
Phase 5 OpenSpec artifacts COMMITTED (
a16c3c5). 1082 LOC. 14 requirements / 30 scenarios. All 5 design questions resolved.Effort
L (~2 weeks)
Dependencies
fulfill-cachedintervention)🤖 Generated by harness-init dogfood + Sisyphus v2 strategy