菜单
turns a raw call, loom, riverside, zoom, or podcast transcript into usable content material. it extracts only the topics with enough substance to become credible articles.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
creates 40 high-intent buyer questions for a specific icp, category, and offer. it finds questions that would be worth answering in a founder or subject-matter expert recording, with a bias toward questions ai engines can answer by citing clear sources.
takes each extracted topic and creates 5 differentiated content angles. each angle is mapped to a buyer segment, trigger, pain, objection, and desired citation format.
selects the best citation-ready article structure from 23 templates and turns an angle into a detailed brief. it optimizes for ai engines that cite direct answers, clear definitions, source-backed claims, and structured evidence.
writes the full article from the optimized brief in the user brand voice. it keeps the useful parts of the source transcript and removes rambling, filler, and fake certainty.
scores the draft from 1-100 and edits it until it is specific, human, useful, and citation-ready. it removes ai slop and flags weak claims before publishing.
prepares and publishes approved content to wordpress, notion, sanity, or a custom cms. it maps metadata, schema, slugs, categories, canonical URLs, and internal links.
基于 SOC 职业分类
| name | transcript-extractor |
| description | turns a raw call, loom, riverside, zoom, or podcast transcript into usable content material. it extracts only the topics with enough substance to become credible articles. |
turns a raw call, loom, riverside, zoom, or podcast transcript into usable content material. it extracts only the topics with enough substance to become credible articles.
raw transcript, the original question list if available, icp definition, and brand voice.
topic cards with claims, proof, examples, objections, quotes, and recommended next agent inputs.
run 01-question-generator first when possible, then record or upload the transcript.
claude "read agents/02-transcript-extractor/SKILL.md and run agents/02-transcript-extractor/prompt.md against inputs/transcripts/<file>.md"
templates/ before producing output.outputs/.