What problem does it solve? Writing a consistent weekly newsletter requires manually reviewing a week of scattered work across meetings, email, GitHub, docs, tickets, and chat sessions, then synthesizing it into a story. This Skill automates that gathering and drafting loop, producing a voice-calibrated newsletter teardown ready for human editing. ## Core Features & Use Cases - Multi-source gathering: Pulls the last 7 days of activity from six sources (meetings CLI, Gmail, GitHub commits across orgs, Google Docs, Linear tickets, pi session JSONL, Slack) in parallel, degrading gracefully when a source is unavailable. - Story filtering and abstraction: Selects the single best teardown candidate using generalizability and business-outcome tests, skipping the week rather than forcing filler content. - Voice-calibrated drafting: Reads accumulated voice lessons from the madcactus MCP server before writing, drafts the six-beat teardown format (tension, what we did, verification, outcome, blueprint, numbers), and lints against voice rules. - Edit-driven learning: After the user edits in the dashboard, reads the draft-to-final diff to derive new voice lessons, and drafts three LinkedIn posts per issue plus a runnable lead-magnet repo folder. - Use Case: A founder asks the agent to draft this week's newsletter; the agent gathers activity from all sources, picks the data-loss-audit story, drafts Issue 02 in the teardown format, pushes it to the dashboard, and the founder edits and schedules it. ## Quick Start Ask the agent to draft this week's newsletter issue from the last seven days of work across all connected sources.
