retell

Transform a Claude Code conversation into a polished blog post through an interactive 5-stage pipeline

Use this agent to create a structured blog post outline from a chosen story angle. This agent reads the full event stream (including relevant subagent content), the author's chosen angle, and any editorial notes, then produces a section-by-section outline with beat treatments, key quotes, and word count estimates. <example> Context: The triage stage is complete, the author has chosen an angle and provided notes user: "Go with the process angle. The brainstorming skill was built in a prior session." assistant: "I'll create a structured outline for the process angle, incorporating your context." <commentary> Stage 3 of the retell pipeline. The outline architect receives the chosen angle, author context, and the full event stream (with relevant subagent content loaded). It structures the post without writing prose. </commentary> </example> <example> Context: Author wants a different structure after seeing the first outline user: "Move section 3 before section 2 and make the rejection a flashback" assistant: "I'll restructure the outline with your changes." <commentary> The outline architect can revise outlines based on author feedback before handing off to the draft stage. </commentary> </example>

Use this agent to assess whether a parsed conversation is blog-worthy and to generate story angle recommendations. This agent reads events.json (the extracted signal from a Claude Code conversation) and outputs a structured triage report with angles, a recommendation, a timeline, and context questions for the author. <example> Context: The retell command has run the parser and produced events.json and manifest.json user: "Turn conversation 8c439a20 into a blog post" assistant: "The parser extracted 82 events. Let me analyze the conversation for story angles." <commentary> Stage 2 of the retell pipeline. The triage agent reads the main conversation signal (no subagent content) and assesses blog-worthiness before committing to the expensive outline and draft stages. </commentary> </example> <example> Context: User wants to know if a conversation has a good story in it user: "Is this conversation worth writing about?" assistant: "I'll use the triage analyst to assess the narrative potential." <commentary> The triage agent can be used standalone to quickly check if a conversation has blog potential before running the full pipeline. </commentary> </example>

This skill should be used when parsing Claude Code conversation files, reading conversation history, working with ".jsonl" conversation transcripts, extracting signal from conversation data, filtering noise entries, linking subagent files, detecting session boundaries, or understanding the Claude Code conversation storage format. Provides the JSONL schema, entry types, content block extraction rules, user message filtering logic, and subagent linking patterns needed by the retell pipeline.

This skill should be used when writing a blog post from conversation data, identifying story arcs in human-AI collaboration, structuring narrative beats, choosing story angles, handling user quotes, or applying editorial principles to conversation-derived content. Covers first-person voice constraints, five-act structure mapping, beat-to-treatment classification, and blog-worthiness heuristics specific to the retell pipeline.