What problem does it solve?
When multiple validators produce conflicting PASS/FAIL votes for the same journey, teams need a reproducible, evidence-first arbitration process to decide which verdict to trust and whether a re-run or human escalation is required. This Skill provides a disciplined protocol to identify diverging criteria, read and summarize each side's captured evidence, and produce a defensible resolution without modifying validator artifacts.
Core Features & Use Cases
- Divergence Identification: Scans per-criterion tabulation to locate exactly which criteria caused the journey-level disagreement.
- Evidence Inventory & Read-Only Analysis: Loads and summarizes each validator's cited evidence, flags missing or zero-byte files, and preserves validator independence by never modifying validator directories.
- Comparative Sequential Analysis: Invokes sequential-analysis with FAIL as the symptom and PASS evidence as the hypothesis seed, running a four-phase investigation and producing a root-cause style report.
- Resolution Emission: Produces one of three exhaustive resolutions (MINORITY_CORRECT, MAJORITY_CORRECT, UNRESOLVABLE) and writes ordered artifacts under e2e-evidence/consensus/disagreement-analysis/ for downstream synthesis or human escalation.
- Use Case: Resolve a split where two validators report FAIL and one reports PASS for a session-persistence criterion by comparing timestamps, screenshots, and network captures and then emitting a criterion-level resolution.
Quick Start
Run consensus-disagreement-analysis on the target journey to compare validator evidence and produce a per-criterion resolution report.
