Claude Skills for Founders: 5 to Upload Today

A founder writes the same five documents on a loop: the requirements for the next feature, the argument about what to cut, the synthesis of what users said, the teardown of a competitor, and the monthly update to investors. Each one has a shape you have to re-explain to Claude every time. A skill is how you explain it once. This guide covers what a skill is, how to upload one to claude.ai in five steps, and the five we built for this job.

What a skill is, in one paragraph

A Claude skill is a folder containing a SKILL.md file: a few lines of metadata (a name and a one-sentence description of when to use it) followed by Markdown instructions Claude follows once the skill is loaded. At the start of a session Claude reads only the descriptions of every installed skill, so twenty of them cost roughly a screenful of text. When your request matches one ("synthesize these interviews"), Claude loads the full instructions and works through them like a runbook. The same file format runs on claude.ai, in Cowork, in Claude Code, and, because Anthropic released it as an open standard, in a growing list of other tools. The complete explanation, including how skills differ from CLAUDE.md and plugins, is What Are Claude Skills?.

How to upload a skill on claude.ai

Anthropic's help center lists skills as available on Free, Pro, Max, Team, and Enterprise, with one prerequisite: code execution has to be on. The steps, verified against the current article as of September 2026:

1. Open Settings > Capabilities and enable Code execution and file creation. On Team and Enterprise plans an owner must also enable Skills in the organization settings.
2. Put the SKILL.md into a folder named after the skill (prd-writer/SKILL.md) and zip that folder so the folder is the root of the archive. The most common upload error the article lists is "Missing required skill.md file," which almost always means the ZIP has an extra layer of folder around it.
3. Go to Customize > Skills, click +, then + Create skill, then Upload a skill, and choose the ZIP.
4. Toggle the skill on with the switch next to it. Disabled skills are invisible to Claude.

Skills do not sync between surfaces. A skill on claude.ai is not in Claude Code and vice versa; each gets its own copy. The claude.ai and API side, including the /v1/skills endpoint for developers, is in Skills on claude.ai and the API. The Claude Code side, where a skill is a folder at .claude/skills/<name>/ in a project or ~/.claude/skills/<name>/ for every project, is in How to install Claude skills. If you have never opened Claude Code, the non-developer walkthrough is Claude Code for Non-Developers and Founders.

Every skill below is pure instruction. No scripts, no network, no repository. That is deliberate: it means the identical file works in all three places, and it means the whole skill is the text you can read on its page before you upload it.

The five skills

1. prd-writer

prd-writer turns a rough feature idea, a Slack thread, or a page of notes into a structured product requirements document: problem, target user, success metrics, scope, non-goals, open questions. The non-goals are the part founders skip and engineers need most.

Use the prd-writer skill. Here is what I know about the feature:
[paste notes]. Our users are independent bookkeepers. The metric we
care about is time-to-first-reconciliation. Flag anything I have not
specified as an open question rather than guessing.

2. mvp-scope-cutter

mvp-scope-cutter takes a feature list and argues it down to what must ship first, with a reason for each cut and the assumptions the remaining scope depends on. Paste it the PRD from the skill above and it will push back.

Use mvp-scope-cutter on this feature list: [paste]. We have one
engineer for six weeks and need something a paying customer can use
by the end. Be aggressive; I will argue back.

3. user-interview-synthesizer

user-interview-synthesizer turns a pile of interview notes or transcripts into a synthesis: recurring themes, supporting quotes, contradictions between participants, and what to ask next time. Its job is to separate what people said from what you hoped they said.

Use user-interview-synthesizer. Here are notes from eight interviews
with clinic managers: [paste]. Tag each theme with which interviews
support it, quote people verbatim, and do not promote anything only
one person said into a theme.

4. competitor-teardown

competitor-teardown produces a structured breakdown of a rival product from whatever you give it: their pricing page, onboarding screenshots, changelog, reviews. Expect positioning, who they are built for, the pricing model, what they do better, what they do worse, and where the gap is.

Use competitor-teardown on Acme. Here is their pricing page text,
their last six changelog entries, and twenty G2 reviews: [paste].
We sell to the same buyer at half the price. Mark anything you are
inferring rather than reading as an inference.

5. investor-update-writer

investor-update-writer drafts the monthly update from your numbers and a few bullet points: headline metrics with month-over-month change, what worked, what did not, the specific asks, and a closing line, in the plain, honest-about-bad-news tone investors actually want.

Use investor-update-writer. MRR 41,200 (up from 37,900), 3 churned
logos, 2 new hires, runway 14 months. We missed the integrations
launch by three weeks. Asks: intros to fintech CFOs. Draft it.

For Claude Code users: commands and an agent

Inside Claude Code, two of these jobs are also slash commands, which is faster when the input is already in the project folder. /prd writes a PRD into the repo from a feature name or an issue, and /scope-mvp reads an existing PRD or feature list from the folder and produces the cut. Both are Markdown files you drop into .claude/commands/.

The technical-cofounder agent is different in kind. It is a subagent, a specialist Claude delegates to with its own instructions and a clean context, built to answer the engineering questions a founder cannot judge alone: is this one service or two, should we buy or build the auth, what breaks first at ten times the load. Ask it before you commit an engineer to a direction, and keep its answers in the project for the engineer you eventually hire. The differences between skills, commands, and agents, and when each is the right shape, are in Skills vs Agents vs Commands.

Making them yours

All five are starting points. The fastest improvement is to add your own template under the skill's instructions: the PRD sections your team actually uses, the investor update format your lead wants, the interview question bank you always run. Edit the SKILL.md, re-zip, re-upload. The third time you find yourself correcting the output the same way, that correction belongs in the file. When you are ready to write one from scratch, Writing Your First Skill is the walkthrough, and the rest of the founder toolkit is at the founders hub.