novusedge/readable-responses/de-slopify

Claude Code and agent skill by novusedge/readable-responses

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when writing, revising, or reviewing technical docs, specs, READMEs, procedures, or troubleshooting guides, or when asked to remove formulaic or AI-sounding prose from blogs and portfolio pages. Not a default style for unrelated conversation or creative writing.

Install

npx vskill@latest install novusedge/readable-responses/de-slopify

SKILL.md

De-slopify

Make writing precise and readable while preserving the author's meaning and voice. Default to targeted edits. A clear sentence may need no change. A wording pass is not a summary; preserve meaningful arguments, examples, and the connections between them.

Choose the task

Honor the requested action: a review produces findings; a rewrite changes the requested text; a wording-only edit preserves structure and scope. Infer the audience and document type from context. Ask only when a missing answer would materially change the result. Project instructions take precedence.

  • Engineering spec: optimize for independent implementation and review.

Read the spec checks in review-checklist.md.

  • User documentation: optimize for completing a task or finding an answer.

Read the user-documentation checks in that reference.

  • Short prose: edit directly. Preserve meaning and omit unrelated checklists.
  • Personal blogs and portfolio pages: retain the author's actual experiences,

opinions, jokes, and rhythm. Distinguish proposed, implemented, and tested work. Use the passage-level checks below rather than imposing a documentation voice.

Editing method

  1. Establish the facts. Read the requested material and relevant definitions.

For technical documentation work, check commands and defaults against available source, help, or tests. For a wording-only request, use the supplied facts. Treat a design spec as proposed behavior: an implementation difference is not automatically a spec error.

  1. Preserve meaning. Keep identifiers, syntax, defaults, conditions, uncertainty,

and requirement strength. Retain useful examples, tables, rationale, and domain vocabulary. Keep technical terms consistent instead of rotating synonyms for variety.

  1. Remove friction. Lead with the outcome or rule. Replace vague claims,

ceremonial introductions, repeated conclusions, and marketing language with concrete facts. Expand shorthand where it conceals behavior. Split a dense bullet when its requirements need separate interpretation or tests.

  1. Check the contract. Compare examples, errors, tables, and tests with the prose.

Resolve a contradiction only when the source makes the intended rule clear. Otherwise record a specific open question and continue independent edits.

  1. Compare the revision with the original. Check for lost conditions, changed

quantifiers, invented behavior, and unnecessary expansion. Report what was actually verified; source inspection is not a live execution test.

Formulaic prose

Read the surrounding argument before editing individual sentences. Look for staged revelations, announcements of honesty, praise for the author's own skepticism, canned questions and answers, and repeated slogan endings. Keep the actual correction, evidence, or implication while removing empty ceremony. A real change of mind may be central to the argument; retain the previous claim and the evidence that changed it when supplied. An opener such as "there goes a line I'd been leaning on" alone supplies neither; cut it without inventing the missing belief.

Preserve good existing jokes, profanity, contractions, and varied sentence lengths. Do not manufacture a human voice with new slang, clipped fragments, deliberate mistakes, or a stock cynical persona. Never invent first-person experiences, motives, feelings, or opinions on the author's behalf.

For example, "The honest version is better anyway: on the matched pairs, the model scored 0.751" can become "On the matched pairs, the model scored 0.751." Keep any substantive correction in the surrounding passage. Conversely, "I spent forty minutes debugging a circuit that was unplugged" may need no edit.

A word or lint match is a prompt to inspect context, not proof of AI authorship or bad prose. Passive voice, a useful contrast, a short paragraph, and an ordinary adjective can all do useful work. Do not replace a promotional claim with a quieter unsupported claim. For a wording-only edit, flag factual doubts separately rather than silently changing the facts.

Read the revision as a passage, then compare it with the original. Restore arguments or examples cut only for brevity, and repair broken connections. Check whether the result still sounds like this author rather than uniformly terse or polished assistant prose. There is no target reduction percentage.

Output

For a rewrite, return the revised text or requested file edits. Add only material unresolved questions and verification limits. For a review, give the location, specific issue, reader impact, and suggested correction; separate editorial fixes from decisions the author must make. If the text is already clear, say so briefly.

Boundaries

The name is not a word blacklist. Do not impose sentence-length caps, banned punctuation, fixed headings, or a tutorial structure on every document. Resolve ambiguity without flattening useful voice or teaching experts basic vocabulary. Never invent commands, defaults, recovery steps, or guarantees to fill a gap. Document review does not authorize running destructive examples or changing code.

This is original practical guidance informed by public plain-language resources and software-documentation concerns. It does not implement or certify ISO 24495-1, ISO/IEC/IEEE 26514, or ASD-STE100. For provenance or standards questions, read sources-and-limits.md.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/novusedge/readable-responses/de-slopify/badge)](https://verified-skill.com/skills/novusedge/readable-responses/de-slopify)

Something wrong with this skill or its page? Tell us.