Brain in the Fish — Document Evaluation

What it does\n\nBrain in the Fish ingests documents (PDF/text), aligns them to evaluation criteria, spawns evaluator agents, and produces evidence-grounded scores and final evaluation reports. It combines ontology-backed agent personas with an Evidence Density Scorer to detect hallucination and provide traceable justifications.\n\n## When to use it\n\nUse when you need rigorous, auditable evaluation of essays, policies, contracts, clinical reports, surveys, or any text where claims must be proven and gaps identified. Suitable for academic marking, policy review, contract checks, clinical governance reviews, and fact-checking workflows.\n\n## What's included\n\n- Scripts: No script folder detected, but the skill documents a set of MCP tool calls (eval_ingest, eval_criteria, eval_spawn, eval_record_score, etc.) for orchestration.\n- References: Detailed workflow, scoring guidelines, supported document types, and architecture notes describing ontologies and evidence scorer.\n- Instructions: Clear orchestration patterns for quick deterministic runs and full multi-agent scoring with debate/challenge cycles.\n\n## Compatible agents\n\nBest used with MCP-style agents that support subagent dispatch and long-running evaluation flows; also compatible with Claude/other multi-turn LLM subagents for scoring tasks.