Writing Reports logo

Writing Reports

Organization
zenobi-us
writing-reports

Writes a self-contained HTML evidence report with screenshots, captions, result badges and a clickable summary table. Use when work must be proved rather than claimed - manual test results, acceptance evidence, an audit, a migration check, a bug reproduction, or any hand-off that needs "here is the screenshot that shows it". Results in one portable directory that renders anywhere, with gaps and leftover test data named.

Overview

Publisherzenobi-us
Repositorydotfiles
Skill namewriting-reports
Stars
67
Forks
6
Bundled files
10
LicenseMIT
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.

  • 10 bundled files

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

  • Open source

    Published by zenobi-us on GitHub. Read the source before you install it.

Installation

Install the Writing Reports 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.
  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/zenobi-us/dotfiles.git /tmp/dotfiles
mkdir -p .claude/skills
cp -r /tmp/dotfiles/files/devtools/agent/bundles/developer/skills/devtools/writing-reports .claude/skills/writing-reports
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Writing Reports 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 Writing Reports 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 Writing Reports 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.

writing-reports

A report is evidence. It states what was checked, shows the screenshot that proves each result, and names what was not covered.

A wall of terminal output is not a report. A Markdown file with no screenshots is not a report. This skill produces a directory that renders in any browser, survives being moved or zipped, and can be attached to a ticket.

When to use

  • Someone asks for a test report, acceptance evidence, QA results, or proof.
  • You finished manual testing and the result must outlive the chat session.
  • An audit, migration, or reproduction needs a durable record.
  • Another skill hands you evidence and a destination. The browser-acceptance-evidence skill does exactly this.

Do not use for a running commentary, a plan, or a design document. A report records what happened, in the past tense, with proof.

Core shape

<destination>/
  index.html              the report
  assets/theme/v1/        tokens.css, base.css     <- never edit
  assets/report/v1/       report.css               <- never edit
  assets/lightbox/v1/     lightbox.js              <- never edit
  shots/                  NN-slug.png
  files/                  companion artifacts the report links to

Every report carries its own copy of the four asset files, linked sibling-relative. That is deliberate: a report gets moved, zipped, downloaded and reopened from somewhere else. A report that only renders inside one directory tree is a report that dies on first contact with a reviewer.

files/ holds the same idea for everything else the report links to: the test plan it ran, the evidence file behind the verdicts, a script or a workflow a reader would re-run. Those artifacts usually live in a sibling directory. A copy goes in files/ so the link survives the move.

Rules

  • You MUST create the report with scripts/new-report.ts. Do not write a report page from scratch, and do not copy the template by hand.
  • You MUST NOT put a <style> block, a style="" attribute, or inline lightbox JavaScript into the page. Every rule lives in the linked CSS.
  • You MUST NOT write a colour into the page. Every colour is a theme token. A raw hex value outside <pre> and <code> fails validation.
  • You MUST NOT edit any file under assets/theme/v1/, assets/report/v1/, or assets/lightbox/v1/ — in a report or in this skill's template. Published reports link those files forever. A breaking change goes in a new v2 directory.
  • You MUST NOT change an asset path to ../ or ../../. The report is self-contained.
  • You MUST copy any companion artifact the report links to — a test plan, an evidence file, a script, a workflow — into <destination>/files/, and link it sibling-relative as files/<name>. You MUST NOT reach a sibling directory with ../, even for a link that is not an asset. A report is handed over on its own; a ../ link dies the first time it moves. Leave the original where it is and treat the copy as the reader's copy.
  • You MUST give every <img> a class, src, alt, width, and height. Without the size the page shifts while screenshots load and every jump link lands on the wrong section.
  • You MUST read each image's real pixel size with scripts/imgsize.ts. You MUST NOT guess a size, reuse the size of another shot, copy the template's placeholder numbers, or trust what a screenshot tool said it produced. The validator reads the file itself and fails on a mismatch. imgsize.ts reads PNG and JPEG headers directly, so it needs no external tool. You MUST NOT write your own size reader, wrapper or shim — if imgsize.ts cannot read a file, save that screenshot as PNG or tell the user, and do not type a number it did not give you.
  • You MUST class every styled node. Strict BEM. A node with no class gets no style.
  • You MUST use the vocabulary in references/markup.md. Do not invent a class. A layout the vocabulary cannot express is a reason to ask the user, not a reason to add a <style> block.
  • You MUST write a "Not covered" section, with an id="notdone", even when it is short. A report that hides its gaps is worth less than one that names them. State the risk each gap leaves.
  • You MUST pass scripts/validate-report.ts before you call the report done.
  • The scripts are TypeScript run by Bun through a mise x -- bun --install=fallback shebang. Run them directly. Do not invoke them with bash, sh, node or python.

CLI

Run scripts/report-cli.ts --help for the Crust router. Use new, validate, imgsize, and doctor. The individual script paths remain direct compatibility entry points.

Procedure

  1. MUST create the report directory:

    sh
    scripts/new-report.ts "<destination-dir>" "<TICKET> &mdash; what this proves"

    The destination must not exist. When another skill supplies a destination, use it verbatim.

  2. MUST save every screenshot into <destination>/shots/, named NN-slug.png. NN orders the shots as the reader meets them, and leaves gaps for later inserts: 10-login.png, 20-banner.png, 31-missing.png.

2b. MUST copy every companion artifact the report will link to into <destination>/files/:

sh
mkdir -p <destination>/files
cp <plan> <evidence> <workflows> <destination>/files/

Link them from the page as files/<name>. A report that points at a sibling directory stops working the moment someone zips it and sends it on.

  1. MUST read the real pixel sizes:

    sh
    scripts/imgsize.ts <destination>/shots/*.png

    It prints <name> <width> <height> per file, read from the file header.

  2. MUST write the result summary first, before any section. One row per check. Every cell of a row carries the same <a class="summary__link" href="#section">, which makes the whole row clickable with no JavaScript. Drop one anchor and that cell stops working.

  3. MUST give every section an id that a summary row targets. A summary row with no target is a broken promise.

  4. MUST pick the badge that tells the truth:

    BadgeUse for
    badge--passThe check passed.
    badge--failThe check failed. Say who owns the failure.
    badge--infoNot testable, not reached, or a neutral finding.
    badge--warnA bug found on the side, unrelated to the change.

    A check you could not run is badge--info, never badge--pass. "The code looks right" is not a pass. A pass needs a screenshot.

  5. SHOULD write a "Test data left behind" section when the work changed shared state: accounts, records, config overrides, running servers. Say what to delete and how.

  6. MUST validate:

    sh
    scripts/validate-report.ts <destination>/index.html

    Fix every line it prints. The script is the gate, not a suggestion.

  7. SHOULD open the page and click one screenshot, to confirm the lightbox opens and the caption reads correctly.

Common mistakes

MistakeWhy it breaksFix
Guessing width/heightThe page reflows on load; jump links land wrongRun scripts/imgsize.ts
A <style> block "just for this one bit"The report stops matching every other reportUse a vocabulary class
badge--pass on an unproven checkThe report lies, and lies are worse than gapsbadge--info plus a "Not covered" row
Caption repeats the alt textalt describes the picture; the caption says what it provesWrite two different sentences
No "Not covered" sectionThe reader assumes full coverageAlways write it
../plan.md in a linkThe report dies when moved or zipped on its ownCopy into files/, link files/plan.md
Editing report.css to fix one pageRestyles every past reportNew v2 directory, or ask

Reference

  • scripts/imgsize.ts — real pixel sizes from the file header.
  • references/markup.md — the class vocabulary, with an example of each block.
  • assets/report-template/index.html — the skeleton, with slot comments.
  • assets/report-template/assets/theme/v1/tokens.css — every colour, size and radius available.
  • assets/report-template/assets/lightbox/v1/lightbox.js — the header comment lists every data-* option the script tag accepts.

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 Writing Reports AI skill do?

Writes a self-contained HTML evidence report with screenshots, captions, result badges and a clickable summary table. Use when work must be proved rather than claimed - manual test results, acceptance evidence, an audit, a migration check, a bug reproduction, or any hand-off that needs "here is the screenshot that shows it". Results in one portable directory that renders anywhere, with gaps and leftover test data named.

Why use Writing Reports on TypingMind?

Because you install it once and use it with any model. Writing Reports 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 Writing Reports in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/zenobi-us/dotfiles/tree/master/files/devtools/agent/bundles/developer/skills/devtools/writing-reports. 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 Writing Reports?

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 Writing Reports?

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

Is the Writing Reports AI skill free?

Yes. It is published on GitHub by zenobi-us under the MIT 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 👇