Evaluate use-self persona debate outputs across three phases and five scoring dimensions.
npx skills add https://github.com/YIKUAIBANZI/forge-skill --skill eval-debate
Please help me install this Agent Skill. Skill: eval-debate Source: https://github.com/YIKUAIBANZI/forge-skill/tree/main/evals/eval-debate Command: npx skills add https://github.com/YIKUAIBANZI/forge-skill --skill eval-debate
It solves the problem of objectively evaluating whether your use-self “persona debate” output is actually high-quality, internally consistent, and aligned with the target persona’s language style.
Ask the assistant to run “/eval-debate” to generate a scored evaluation report for the specified persona and its decision scenarios.
High-intent search queries and answers about installing and using this skill.
To evaluate persona-driven decision debates, load YAML test cases and persona.json files to run a three-phase workflow of independent analysis, challenge, and synthesis, scoring five dimensions per scenario.
Persona evaluation for prompt benchmarking is the process of scoring a persona's debate outputs against expected stances across multiple decision scenarios to measure variant distinctness, challenge depth, and parameter consistency.
You score multi-phase debate outputs by applying a five-dimension rubric covering variant distinctness, challenge depth, parameter consistency, synthesis coverage, and user language style across test cases loaded from YAML files.
Yes, you can run local-first debate evaluation entirely within the current conversation without external API calls, completing the three-phase analysis and generating scored reports locally.
Yes, YAML test cases and a persona.json file are required to benchmark persona debate scenarios, providing the ground truth parameters and decision scenarios needed for the evaluation workflow.
The best way to assess language style consistency is to run a persona-grounded evaluation that scores the user language style dimension alongside variant distinctness and synthesis coverage across multiple decision scenarios.
EntityProcessCreate and manage AgentV YAML evaluation files for AI agent performance testing.
cagdotinAutomate drafting, evaluating, and iterating Claude Skills with SKILL.md.
fockusCoordinate three judge agents in iterative debates to produce consensus evaluations.
SimHackerCreate weighted rubrics and score outputs against defined criteria.
