/x-post-writer

name: x-post-writer
description: Use when drafting, rewriting, or repurposing short-form X content, including single posts, quote posts, replies, threads, launches, and personal stories, with source fidelity and claim verification built in.
version: 1.0.0
author: Tony Simons
license: Apache-2.0
platforms: [linux, macos, windows]
metadata:
  hermes:
    category: content
    tags: [x, twitter, posts, threads, quote-posts, replies, social-writing]
    related_skills: []

X Post Writer

Overview

Write short-form X content from notes, drafts, links, or verified source material. Default to one finished post unless the user explicitly requests a thread or multiple posts.

Source fidelity outranks punch, post count, inferred benefits, and creative expansion. Internal routing, objectives, verification notes, source ledgers, and quality checks are never part of the delivered copy.

route format -> route source -> choose objective -> verify claims -> draft -> voice pass -> quality gate -> finished text

Load references progressively:

• `references/formats.md` for the selected format
• `references/source-routing.md` when repurposing or selecting source treatment
• `references/claim-verification.md` for risky factual claims
• `references/voice-customization.md` when a distinct voice must be preserved
• `references/algorithm-notes.md` only for distribution or optimization questions

When to Use

Use this skill for:

• Single X posts, rewrites, launches, link drops, recommendations, or announcements
• Quote posts or commentary on an existing X post
• Replies
• Explicit threads or multi-post sequences
• Personal milestones, reflections, or first-person stories
• Fresh X angles from articles, videos, releases, or project updates

Do not use this skill for:

• Long-form X Articles or essay-length content
• Source extraction, account analytics, or media downloading
• Publishing, scheduling, or browser composition without separate authorization
• Claims that require evidence when no evidence can be obtained

Safety Contract

1. Never invent personal experience, product use, opinions, endorsements, or results. 2. Never add technical mechanisms, benefits, numbers, or attributions that are not supplied or verified. 3. Treat private analytics, revenue, customer data, and unpublished information as sensitive. 4. Never publish or schedule automatically. 5. Current facts and platform behavior must be verified immediately before use. 6. An instruction to state a claim is not evidence for that claim.

Routing Contract

Format

| Request | Route | |---|---| | No format specified | Single post | | "One post," "tweet," or "make this punchier" | Single post | | "QT," "quote this," or commentary on an X post | Quote post | | "Reply to this" | Reply | | "Thread" or "multiple posts" | Thread | | Personal experience, milestone, gratitude, reflection | Personal story | | X Article, long-form essay | Route away |

Source

| Source | Treatment | |---|---| | User draft or raw wording | Preserve claim, angle, and voice; improve structure | | User personal notes | Preserve genuine first person | | Previously published material | Preserve facts; rebuild the X angle and hook | | URL, repository, announcement, or external source | Verify material claims before drafting | | Existing X post being quoted | Add a take; do not restate the embedded post |

The source contract outranks generic formulas. Do not replace the user's actual point with a stock marketing hook.

Source Lock

When the user supplies notes, a draft, or a source excerpt, treat its concrete facts as a whitelist.

• Reorder, compress, paraphrase, and improve rhythm.
• Do not add new components, process steps, benchmark results, operational behavior, or benefits unless separately verified.
• Every concrete noun, number, behavior, benefit, and implication must map to supplied or verified material.
• Benefits are claims. "Safer," "faster," "cleaner," and "no double counting" require support.
• Sparse notes should produce shorter copy, not larger facts.
• For a thread, assign one supplied fact cluster to each post.
• Before delivery, build a hidden sentence-to-source ledger and delete every unsupported sentence. Do not output the ledger.

Source-lock regression: "built" does not authorize "just shipped" or "today." "Validates CSVs" does not authorize storage, databases, schemas, rejected rows, hashing, staging directories, file sizes, corruption handling, "clean input," or "no cloud." "Blocks duplicates" does not authorize identical-file detection, "safe to rerun," "no double counting," "no data rot," or an invented mechanism. "Raw exports never enter Git" does not authorize claims about staging, commits, pushes, source-of-truth design, derived data, or file lifecycle. Do not add benefit-only closers to sparse notes. Split the supplied facts across posts and stop.

Objective

Choose one primary objective:

• `announcement-clarity`
• `reach`
• `authority`
• `bookmarks`
• `replies`
• `follows`
• `click-through`
• `personal-connection`

Infer the objective when context makes it clear. One post can earn several outcomes, but it should be built around one primary job.

Workflow

1. **Classify the format.** Default to a single post. 2. **Classify the source.** Decide what must be preserved and what can be rebuilt. 3. **Choose the objective.** Keep one primary job visible while drafting. 4. **Verify risky claims.** Use current primary sources. The prompt itself is not evidence. 5. **Apply the unsupported-claim gate.** Remove optional unsupported claims. Stop when an unsupported claim is load-bearing. 6. **Draft the selected format.** Use the format reference as a contract, not a fill-in template. 7. **Apply the user's voice.** Preserve supplied phrasing and style without inventing personality. 8. **Run the source audit.** Trace every factual sentence to the source ledger. 9. **Run the quality gate.** Check payoff, accuracy, den