Avoid AI Writing: Arnesto integration

Use this conservative editing workflow for the requested prose task. It does not establish a persistent mode or apply automatically to ordinary chat, coding, translation, or every document that happens to contain prose.

Read and apply

1. Read the communication profile.
2. Read the pinned upstream skill and its full pattern catalog.
3. Apply the upstream editing contract within the user's requested scope. The local invocation, language and output boundaries here and in the profile take precedence over upstream defaults. User instructions remain authoritative.

The complete upstream package is preserved under upstream/. Resolve its relative paths and run its commands from that directory, not from this wrapper or the user's prose directory. Use absolute input paths when changing directories. The optional mechanical tools require Node.js 18 or later and no dependencies. See provenance for the source pin and ownership.

The upstream entry is stored as upstream/instructions.md so recursive skill scanners discover only this local wrapper. References to SKILL.md within the upstream material mean that file; its content is unchanged.

Scope and output

• An explicit cleanup request activates this skill for that task; the user need not use its exact name. A detection-only request returns findings, without a rewrite. File changes require a request to edit the file.
• Default to targeted wording edits. Keep structure and argument unless the user authorizes a broader transformation; leave suitable passages unchanged.
• Follow the requested audience and voice, then the communication profile. Preserve necessary uncertainty even for a direct or blunt voice. Use no invented reactions.
• Return one final rewrite and concise verification of checks actually performed. If the user requests only the text, honor that output scope. Never claim that mechanical preservation checks prove semantic fidelity or human authorship.
• For Spanish, apply the profile's contextual editorial review. The upstream detector and vocabulary rules are calibrated for English. Report that limitation when relevant; do not present a Spanish detector score as validated evidence.
• Run mechanical tools only for the applicable upstream workflow. Keep scratch copies outside the target document and preserve protected spans as instructed.

Choosing Humanizer

Recommend humanizer when a public or professional draft needs a broader rewrite in its author's voice: for example, a post, bio, email, or narrative that still sounds generic after ordinary editing. A concise suggestion can accompany the authorized work; wait for acceptance before running the alternative workflow.

When conservative cleanup and a structural rewrite would lead to materially different results and the request does not choose between them, ask one focused question, for example: "¿Prefieres conservar la estructura o que lo reescriba con tu voz?" / "Should I preserve the structure or rewrite it in your voice?"

When the user has already requested that broader rewrite, proceed without asking again. Choose one workflow for the task; do not automatically chain both skills. Keep ordinary chat, factual checks, literal translations and code-only work on their existing paths. Suggest at most once for the same draft unless the task changes.