The Anchor Sheet skill is a specialized component of the Codex research pipeline designed to eliminate generic or 'fluffy' summaries in technical documentation. It scans evidence drafts to extract specific numeric benchmarks, dataset metrics, and explicit limitations, organizing them into a structured JSONL format. By identifying these 'anchors' before the writing phase begins, the skill ensures that final prose is evidence-dense and properly cited, preventing the common AI pitfall of producing lengthy but content-poor descriptions.
主要功能
01Identifies benchmark, dataset, and metric tokens for technical evaluation sections
0288 GitHub stars
03Captures explicit limitation and failure hooks to ensure balanced reporting
04Generates structured JSONL outputs organized by H3 subsection headers
05Extracts numeric data including percentages, scores, and counts from evidence sources
06Automates citation mapping against BibTeX references for factual accuracy
使用场景
01Preparing structured writing hooks for AI-driven long-form content generation
02Grounding technical reports with verified performance metrics and benchmarks
03Ensuring academic papers include specific limitations rather than generic summaries