What problem does it solve?
This Skill verifies that a research step’s stated truths, artifacts, and evidence actually match the published or local inputs so you can trust downstream decisions.
Core Features & Use Cases
- Truths validation against evidence: Checks that each declared plan Truth is observable in the target content for default research (published URL or local file).
- Receipt verification by learning variant: Confirms GitHub rejection/calibration issue content matches the draft reports (with normalization) and that dataset artifacts are consistent with their manifest.
- Operator-gated failure modes: Converts missing files, non-200 publications, drift, and schema mismatches into clear gate requests with recommended remediation.
Quick Start
Use the verify/research skill to validate a playbook: research action by checking its external_refs (published URL, local file, GitHub issue receipts, or dataset artifacts) and producing a verified/partial/failed verdict.