What problem does it solve?
Manually evaluating partnership pitch angles for factual accuracy, differentiation, and persuasiveness is time-consuming and prone to oversight, risking the use of fabricated or weak content in prospect-facing videos that could damage brand credibility. This skill automates the quality evaluation process, applying consistent, rigorous grading criteria to catch issues before production.
Core Features & Use Cases
- Four-Dimension LLM-as-Judge Grading: Evaluates angles across grounding (40% weight), distinctness (20%), capability fit (25%), and persuasiveness (15%) with detailed rubric anchors.
- Hard Guardrails Against Fabrication: Automatically fails evaluation and flags BLOCKER concerns if any fabricated or uncited claims are detected in angle beats, preventing false information from reaching prospects.
- QA Gating: Skips full grading if upstream partnership-angles inline QA fails, avoiding wasted effort on invalid artifacts.
- Structured Verdict Output: Generates a standardized YAML verdict with dimension scores, overall disposition, and auto-surfaced concerns for downstream aggregation by the opp-eval skill.
Use Case: ACE partnership teams building prospect-facing video pitches can use this skill to automatically validate generated partnership angles, ensuring they are factually grounded, differentiated, aligned with real Connect capabilities, and persuasive enough for production.
Quick Start
Invoke the partnership-angles-eval skill to automatically grade the partnership angles artifact for the active prospect run and produce a structured verdict YAML.
