Extend watch lists with recycled emphasis and hook tells

[Scott-Emberson]

Extends the watch lists of five existing patterns with phrases that show up constantly in AI-assisted social and blog writing but are not yet named in the skill. No new pattern numbers, no renumbering; this follows the shape of the "quietly" (#192) and §28 casual-register (#219) additions.

• §31 (forced punchlines): adds a Words to watch line for manufactured gravity: "Read that again", "Let that sink in", periods between single words ("every. single. day."), and single ALL-CAPS words for emphasis. All four command a dramatic pause the claim has not earned, and §31 previously had no watch line to anchor them.
• §33 (fake-candid openings): adds "I'm no expert but" and "This might be controversial but". Both are staged modesty before a confident, usually uncontroversial claim, the same move as "Honestly?".
• §28 (announcing the next point): adds "what if I told you" and "here's what nobody tells you". Hook templates that promise an insight instead of stating it.
• §27 (fake deeper truth): adds "the truth is", which front-loads candor before an ordinary claim.
• §18 (emojis): notes that decorative arrow chains (→) used as a mock flow chart are the same decoration problem; write the steps as a sentence.

Source: comparative review against a widely circulated third-party "humanizer" prompt whose pattern list is built from ongoing LinkedIn feed analysis, cross-checked against posts in the wild. Only phrases that read as recycled templates in general prose made the cut; platform-specific advice (engagement mechanics, algorithm claims) was deliberately excluded as out of scope for this skill.

Checks: validator passes ("Humanizer package v2.11.2 is valid"), SKILL.md is 458 lines (limit 500), patterns 1-35 unchanged and sequential. README pattern-table cells are untouched since they illustrate rather than enumerate; happy to update them if you prefer. No version bump to avoid colliding with the other open PRs.

🤖 Generated with Claude Code