What problem does it solve?
Partnership research artifacts (deep web research reports and Connect/Dimagi capability-fit memos) often contain sourcing gaps, factual errors, or content misaligned with prospect expansion goals, leading to low-quality or risky prospect-facing materials. This skill provides consistent, objective LLM-as-judge quality grading to catch these issues before artifacts are used in partnership pitches.
Core Features & Use Cases
- Four-dimension LLM-as-judge grading: Evaluates research artifacts across grounding, relevance, capability fit, and factual safety using a standardized rubric with hard deduction rules for critical issues like fabricated claims.
- QA gating: Automatically skips evaluation if prior partnership-research-qa checks failed, avoiding wasted effort on invalid or incomplete artifacts.
- Standardized verdict output: Writes a structured YAML verdict with dimension scores, weighted overall scores, and auto-surfaced severity concerns for aggregation into opportunity evaluation workflows.
Use case: For a partnership development team building pitches for Connect opportunities, this skill automatically evaluates the quality of deep research and capability fit memos to catch unsourced claims, generic content, or factual errors before sharing materials with prospects.
Quick Start
Use the partnership-research-eval skill to evaluate the quality of the partnership research artifacts for the current prospect and generate a standardized quality verdict YAML.
