TL;DR: The mock-reader-feedback skill is a Claude Code sub-command that uses 5 preset personas (P1-P5) to simulate reader feedback. It outputs a YAML feedback report to docs/feedback/<article>-<persona>.md. This article breaks it down: how the 5 personas are constructed, how the 7-step workflow runs, how the YAML schema is strictly enforced, and 5 anti-patterns I have already hit. Author’s note: This skill is what I crystallised after running 12 feedback reports against 3 articles in August 2026 (hugo-cloudflare-pages-pitfalls / claude-code-cli-setup-indie-blog / claude-code-editorial-pipeline). Every report, script, and persona file referenced here lives in the repo and is inspectable. ...