From amplify
Removes AI writing patterns like filler phrases, formulaic structures, and robotic rhythm from prose to sound human. Activates for editing AI-sounding text, skipping code or grammar.
npx claudepluginhub wunki/amplify --plugin ask-questions-if-underspecifiedThis skill uses the workspace's default tool permissions.
Eliminate predictable AI writing patterns from prose while preserving the author's voice and intent.
Rewrites text to remove 24 AI writing patterns like inflated language, em dashes, filler phrases, and formulaic structure, making it sound natural and human-written.
Reviews and edits copy to remove AI-generated patterns using 30 rules, making text sound natural. For editing drafts, reviewing robotic copy, or cleaning AI text.
Removes AI-generated writing patterns from text and injects personality to sound natural and human-written. Use after drafting docs, emails, or copy.
Share bugs, ideas, or general feedback.
Eliminate predictable AI writing patterns from prose while preserving the author's voice and intent.
Read references/phrases.md when identifying or removing filler phrases, business jargon, or emphasis crutches. Skip if the user's request is purely structural (rhythm, fragmentation).
Read references/structures.md when identifying or fixing binary contrasts, dramatic fragmentation, rhetorical setups, or rhythm problems. Skip if the user's request is purely about specific word choices.
Read references/examples.md when the user asks for examples, before/after rewrites, or when demonstrating a transformation would clarify a correction. Skip if the user has not asked for examples and the correction is self-evident.
Cut filler phrases. Remove throat-clearing openers and emphasis crutches. Specific phrases are in references/phrases.md.
Break formulaic structures. Avoid binary contrasts, dramatic fragmentation, rhetorical setups. Patterns are in references/structures.md.
Vary rhythm. Mix sentence lengths. Two items beat three. End paragraphs differently.
State facts directly. Skip softening, justification, hand-holding.
Cut quotables. If it sounds like a pull-quote, rewrite it.
Preserve voice. Remove patterns without flattening the author's distinctive choices. A phrase is deliberate (not slop) when it appears consistently in multiple places, uses unusual word choice unique to the author, or is structurally necessary (e.g., a technical instruction like "Navigate to Settings"). Flag it rather than silently removing it.
Skip technical content. Do not apply prose rules to code blocks, filenames, CLI commands, UI labels, or technical terms embedded in prose (e.g., "Navigate to the Settings page" in a tutorial is a UI instruction, not business jargon).
Before delivering prose:
Rate 1-10 on each dimension:
| Dimension | Question |
|---|---|
| Directness | Statements or announcements? |
| Rhythm | Varied or metronomic? |
| Trust | Respects reader intelligence? |
| Authenticity | Sounds human? |
| Density | Anything cuttable? |
Below 35/50: revise. Perform one revision pass, then re-score. Stop after the second pass regardless of score — additional passes risk changing the author's meaning. If the score is still below 35 after two passes, note the remaining issues and ask the user whether to continue.
If the text is under 50 words, skip scoring and apply the Core Rules and Quick Checks only.
Determine the task:
If the text is ambiguous in intent (e.g., a stylistic fragment that could be deliberate), flag it with a one-line note rather than silently removing it.
Deliver the output. If patterns were intentionally left, list them at the end with a one-sentence reason for each.