Open source · Free, MIT-licensed
Everything you ship gets read cold — by followers, users, customers, or your board. ThoughtfulBits Skills reads it cold first: a fast post edit, a screen-by-screen UI test, a SPARK score for the next feature. And when board season comes, the deep deck audit.
v1.4.0 · Six skills · No ThoughtfulBits server
post-editor
Voice kept
“Your story is in line 4 — a cold reader never gets there. Lead with the number.”
test-ui-ux
Score 6.8 / 10
“Inviting a teammate takes 9 steps and 6 fields. The simplest safe path is 3 steps — prefill the rest.”
product-feature-feedback
SPARK 19 / 25
“Purposeful: 5. Known: 2. It solves a real problem — and nobody will find it. Fix discovery first.”
The flagship · board season
The skill that started this package: eight checks applied the way a tough director reads, a verdict, and the 3–7 fixes that matter — while there's still time to use them.
Read the eight checksboard-deck-audit
Ready after small fixes
01 · The checklist
01
Flags every buzzword — “best-in-class,” “hypergrowth flywheel” — and rewrites it in plain words. If a phrase could appear in any company's deck, it carries no information.
02
Can the strategy be stated in one meaningful sentence — the customer pain, and how you solve it? If the deck doesn't say it, that becomes fix #1.
03
Every roadmap item should answer: how many customers did we talk to, and what did we learn? Analyst TAM projections don't count.
04
What are competitors doing, and are we winning or losing against them — with win/loss data. “We have no real competitors” gets called out.
05
Every section — sales, churn, financials, team — needs the number and the honest cause behind it. “Churn rose because cancellations increased” is a what wearing a why costume.
06
Plan vs. actual on every headline number, consistent metrics with no selective-disclosure games, and closed loops on what you told the board last quarter.
07
Burn, runway, and the next-raise date math stated plainly — the board should never have to do the subtraction themselves.
08
Decisions needed vs. help wanted, each one decidable from the deck — and discussion topics that match the deck's own biggest problems.
02 · The output
A structured pre-review, in chat and as a shareable markdown file: a verdict (ready / ready after small fixes / needs revision), a jargon table with rewrites, section-by-section what/why analysis, an accountability-and-cash readout, the questions the board will ask, and a prioritized fix list of the 3–7 changes that actually matter.
03 · The plugin
Ordered by how often you'll reach for them — the everyday editors first, the deep board audit for when the meeting is on the calendar. Attach your material, point it at a UI, or paste a social draft; ask in plain language, and the right skill triggers on its own.
01
Edits existing short- or medium-form X, LinkedIn, and social-caption drafts. It locks the facts, intent, and voice first, then improves the opening, reader value, concrete proof, emotional recognition, rhythm, shareability, and platform fit — without manufacturing evidence, outrage, or algorithm folklore. Inspired by Nobunaga, Jonah Berger, and Katherine Milkman.
02
Identifies every screen's important actions, counts their steps, clicks or taps, and fields, then recommends the simplest safe path using optional, prefilled, derived, or editable AI-drafted inputs. Five isolated evaluators — SPARK, Nielsen heuristics, cognitive walkthrough, PURE, and WCAG 2.2 AA — still produce the arithmetic-average 1–10 score, critical-failure gate, evidence, prioritized fixes, and loop-ready JSON.
03
Scores a single feature or product on the SPARK method — Simple, Purposeful, Attractive, Reliable, Known — out of 25, with the three simplest improvements to make next.
04
Grades a product plan, roadmap, launch plan, or GTM plan against a key-milestone rubric — strategy, executors and champions, metrics, design partners, positioning, and pricing.
05
The quick read — the concise reaction a sharp, experienced director would give. Answers one question: does this deck make sense? One screen, no tables.
06
The deep pre-send audit described on this page: a claim ledger, reconciled math, evidence grading, and a prioritized fix list.
04 · Install
ThoughtfulBits Skills is distributed through OpenAI's Plugins Directory for ChatGPT Work and Codex, as a public marketplace plugin for Claude Desktop, and as a portable skills.sh repository for compatible agents.
ChatGPT Work & Codex
Open the verified ThoughtfulBits Skills directory listing, select the plus button, and start a new chat. The same installation works in ChatGPT Work and Codex; directory updates may arrive after the GitHub release.
Open the OpenAI listing ↗Supported in ChatGPT Work on web/desktop, Codex desktop, and Codex CLI — not ordinary Chat, mobile, or the IDE extension.
Claude Desktop
Add the public GitHub repository as a plugin marketplace and sync it. The five illustrated steps below take about a minute.
View the Claude install steps ↓No terminal is required.
skills.sh
Use the public repository with the open skills installer. Install post-editor by itself or install all six skills into every compatible agent directory.
Open the install commands ↗No ThoughtfulBits account or server is required.
Installation takes three steps in the OpenAI account or workspace where you want to use the skills.
STEP 01
Open ThoughtfulBits Skills in the OpenAI Plugins Directory while signed in to the account or workspace where you want to use it.
STEP 02
Install the plugin from its listing. ThoughtfulBits Skills contains instructions only — there is no ThoughtfulBits account, server, or separate data connection.
STEP 03
Attach a deck, plan, roadmap, feature spec, or UI screenshot; provide a live URL; or paste a social draft. Type @ThoughtfulBits Skills to invoke it explicitly, or ask in plain language.
STEP 01
Open Claude Desktop. In the left sidebar, choose Customize, then Plugins.
STEP 02
Click Add, then choose Add marketplace.
STEP 03
Choose Add from a repository.
STEP 04
Paste the public repository URL below, then click Sync.
https://github.com/thoughtfulbits/thoughtfulbits-skills
STEP 05
Claude adds and enables ThoughtfulBits Skills. Confirm that its toggle is on.
Test this UI thoroughly. For every screen, identify the actions that matter, count the steps, clicks, and fields, and show the simplest safe path before giving me the five-agent 1–10 score.
Start a new supported ChatGPT Work, Codex, or Claude Desktop conversation, or install through skills.sh. Provide a live URL, screenshot, product spec, attached document, or pasted social draft. Live UI tests can produce a full verdict; screenshot- and spec-only UI reviews remain provisional. AI-generated input is proposed as an editable draft for your confirmation, never silently submitted.
Built from the criteria I use reading real board decks, product plans, feature specs, interfaces, and public writing. MIT-licensed — use it, fork it, and make the next iteration better.