Grounded Writing logo

Grounded Writing

CommunityPopular
chrisbanes
grounded-writing

Use when drafting or reviewing public developer documentation, other text the user will publish or send, or a user-owned draft under an explicit editorial review, including review comments, replies, and internal technical reports.

Overview

Publisherchrisbanes
Repositoryskills
Skill namegrounded-writing
Stars
1.1K
Forks
53
Bundled files
2
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 2 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by chrisbanes on GitHub. Read the source before you install it.

Installation

Install the Grounded Writing AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/chrisbanes/skills/tree/main/skills/grounded-writing
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/chrisbanes/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/skills/grounded-writing .claude/skills/grounded-writing
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Grounded Writing in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Grounded Writing on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Grounded Writing is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Grounded Writing

Core principle

Make the reasoning visible at the scale the artifact supports. Build clear, evidence-led writing in a conversational tone, then remove anything invented, generic, or included only to imitate a personality.

Procedure

  1. Confirm that the text is for the user to publish or send, or that the user explicitly requested an editorial review of their own draft. Apply this style at any length for publication or sending, including one-sentence review comments and replies. For an internal report review, apply the truth and material-edit checks without imposing public-documentation style. Do not apply the skill to an ordinary assistant reply, quoted source text, or prose attributed to someone else.
  2. Read the style profile before drafting or revising.
  3. Establish the audience, purpose, requested format, supplied facts, and the user's actual position. Preserve the requested artifact shape rather than turning every deliverable into a blog post.
  4. In public developer documentation, explain the observable difference between settings and the practical trade-off in supported terms: what each gains and gives up, and when to choose each setting. For a publication review, name these as separate checks even when the draft lacks evidence to fill them in. In a review, explicitly recommend removing each implementation mechanic present in the draft that has no established reader-facing significance; merely identifying it as a mechanic leaves the edit ambiguous. If the draft lacks evidence for that choice, name the missing information instead of inferring it. Remove implementation, test, or diagnostic mechanics such as sampling thresholds, CPU masks, and interpolation details. Retain an observable transition or associated timing only when supplied or verified evidence shows it is part of the public contract or materially informs how to use or choose the setting. If a draft says one setting returns to another, preserve that observable transition conditionally while checking whether its timing is a public contract. Say that it returns to the other setting without carrying over a “cooldown” term or numeric delay unless that timing is verified as reader-facing behavior. Do not remove the return behavior merely because its implementation timing is unverified. Technical specificity alone does not establish public relevance. If relevance could change the recommendation but is unknown, flag it for verification rather than treating it as settled behavior. Describe retained effects in outcome terms and omit their implementation mechanics. Include a mechanic only when it is required for correct API use, necessary to distinguish a reader-facing setting choice, or explicitly requested. Do not ask to expand internal mechanics as a way to fill a public-documentation gap. Before drafting or revising public developer documentation, identify its public contracts and guarantees shared across alternatives, including compatibility expectations, and check that supported guarantees remain explicit in the result. If the source leaves a contract uncertain, verify it when research is in scope or flag the gap rather than silently dropping or inventing it. Retain the context needed to interpret claims. This boundary does not apply to internal design documents or technical reports.
  5. Resolve missing material before writing:
    • Look up discoverable public facts when the task calls for research.
    • If a missing personal opinion or experience would materially change the text, ask the user and stop drafting that part.
    • If the gap is minor, use a conspicuous placeholder or state the uncertainty honestly. Never invent a first-person claim, result, preference, or memory.
  6. Choose the register from the style profile. Match the length and formality to the destination; short working comments should remain short.
  7. Shape the reasoning before polishing sentences. Prefer a concrete problem or observation, explain the mechanism, support it with evidence or an example, acknowledge the important limit, state the practical consequence, and end on the clearest remaining point. Omit any stage the artifact does not need. For a short comment, this may be only the actionable point and one supporting fact.
  8. Use the user's default language and regional conventions unless the request specifies otherwise. Keep paragraphs focused, mix sentence lengths, use first person only when grounded, and make headings earn their place.
  9. Edit once for style and once for truth. Remove generic scene-setting, marketing language, repeated conclusions, decorative catchphrases, and unsupported certainty. Treat the headline and opening claim as substantive: remove or qualify promotional framing that the supplied evidence does not directly support.
  10. For a review that permits edits only for material clarity or truth problems, identify the exact claim that is false or the specific misreading that would prevent the document's intended use before touching the file. If neither is present, leave an adequate draft unchanged. Rewording an already usable instruction to make it more explicit is optional polish, not a material clarity fix. Do not turn an internal report into a rerun protocol: missing identifiers or reproducibility details justify an edit only when a stated claim or action depends on them. Do not add a placeholder for an unverified detail unless it blocks the draft's intended use. Explain the material facts, mechanism, and qualification that make a no-change result appropriate; a bare “no edit needed” does not show that the checks were completed.

Finish gate

Finish only when all of these are true:

  • The result still satisfies the requested format and purpose.
  • For reader-facing setting documentation, a developer can tell what changes, what each setting gains and gives up, when to choose each, the public contract, including supported guarantees shared across options, and the meaningful limitation. Include transitions or timing only when evidence establishes their public or decision relevance; otherwise omit them with the implementation detail or flag material uncertainty. Keep implementation mechanics out unless they are required for correct API use, necessary to distinguish the setting choice, or explicitly requested. In a review, ask separately for the visible difference, the practical trade-off, and when to choose each setting; a request for a trade-off alone does not cover the visible difference. Explicitly recommend removing unsupported mechanics present in the draft.
  • Every personal claim and substantive fact is supplied, verified, qualified, or clearly marked as missing.
  • The argument is concrete enough to follow without promotional filler.
  • Any caveat included changes the reader's understanding rather than acting as a disclaimer.
  • Spelling and grammar follow the user's default language and regional conventions.
  • The ending lands once and does not recap the whole piece.
  • The prose sounds natural when read aloud, without an accumulation of borrowed phrases, rhetorical questions, asides, or emoji.

If a check fails, revise the draft. If the failure depends on an unknown personal position, ask the user rather than smoothing over the gap.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Grounded Writing AI skill do?

Use when drafting or reviewing public developer documentation, other text the user will publish or send, or a user-owned draft under an explicit editorial review, including review comments, replies, and internal technical reports.

Why use Grounded Writing on TypingMind?

Because you install it once and use it with any model. Grounded Writing is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Grounded Writing in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/chrisbanes/skills/tree/main/skills/grounded-writing. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Grounded Writing?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Grounded Writing?

As many as you like. As long as a model supports skills, you can use Grounded Writing with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Grounded Writing AI skill free?

Yes. It is published on GitHub by chrisbanes under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇