new-persona
Skill5 starsChanged 13 days agoApache-2.0View on GitHub
Create one or several personas for the connected app - read the app to find what about a user changes its screens (account state, plan, role, how much data, language), ask the person the questions that define each profile, then save each persona with its own mock data and optional photo, and check the screens as them. Use when the person asks for a persona, a user profile, a test user, "someone who…", or more kinds of users to try the app as.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- connect-appSkill
- mirror-screenSkill
- new-proposalSkill
- update-herocanvasSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 37
One persona: ask what defines it. In Claude Code, with the question tool, up to four questions a round, options taken from the app's own …
- Distinctive: line 59
Never download a photo from the web on your own, and never use a real person's photo without their say: suggest initials or a generated one …
- Very specific: line 76
Look at the screens the persona changes, as them: npm run shot -- "/<page>?persona=<id>", phone and tablet, light and dark.
- Very specific: line 79
Add an entry to the root CHANGELOG.md (and bump package.json's version): it's a change to the team's mirror.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
New personas
A persona is someone the app is tried as: picked in the Scenarios panel (?persona=<id>), every screen shows
their data. HeroCanvas only carries the pick; the app declares them (appConfig.personas: id, name, a one-line
description, an optional avatar) and its mock data module gives each one their data. One module holds it all:
PERSONAS and the data per persona id (the sample-tasks tag has an example, its app/data/sample.ts).
A persona is who someone is, lasting: a new user with nothing yet, a paying admin with a lot. What any user can run into for a moment (loading, an error, offline, a sheet open) is not a persona: it's a page's state or a scenario control in the manifest.
1. Read the app: what about a user changes the screens
In the person's app's code (read, never edited) and in the mirror (app/), find what differs from one user to
another and shows on screen:
- Account state: just signed up (empty states, onboarding not finished), active, lapsed, a trial ending.
- Plan and entitlements: free or paid tiers, feature flags, limits ("3 of 5 projects").
- Role and permissions: owner, member, viewer, admin: what they can see or do.
- Their data: how much (none, one, a lot), what kind, its extremes (a very long name, an overdue item, a big number), what's connected (integrations, a team, a device).
- Where they are: language, region, currency, units, time zone, when the app varies by them.
Tell the person what you found in a few lines: the dimensions, the values the code allows, the personas that exist and what none of them covers yet (no paying user, no one with a lot of data). Don't ask what the code answers.
2. Ask
Several personas: first propose a set that covers the dimensions that matter, as a short table (name, who they are, the values that set them apart), and confirm it. Then ask only what's left open for each.
One persona: ask what defines it. In Claude Code, with the question tool, up to four questions a round, options taken from the app's own values (its plans, roles, states), your recommendation first; other agents ask in plain text. Typically:
- Who: what they use the app for, in a sentence; their name (offer one).
- Account and plan: state, tier, role, from what the app has.
- Their data: how much and what's notable about it, concretely ("two projects, one with an overdue task").
- Language or region, only if the app varies by them.
- Photo: none (their initials), an image they give, or one they generate (step 3).
Names are invented: plausible where the app is used, varied, never a real person's, no stereotypes tied to the profile.
3. The photo
- None (the default): the app's avatar shows their initials. Nothing to do.
- An image the person gives (a path, or one dropped in the chat):
npm run persona:avatar -- <image> <id>crops its centered square, scales it to 256 px and writespublic/personas/<id>.png; put the printed path in the persona'savatar. PNG, or anything macOS'ssipsconverts (JPEG, HEIC). - One they generate (an illustrated avatar, from an image tool of their choice): the same command, once they pass the file.
Never download a photo from the web on your own, and never use a real person's photo without their say: suggest initials or a generated one instead.
4. Save
- The persona: in the app's personas list (the one
appConfig.personasreads), with a kebab-caseid, itsname, adescriptionof one line for the Scenarios panel ("Admin on the paid plan, a busy team"), andavatarif there's a photo. The list's first persona is the default: add new ones after it unless the person says otherwise. - Their data: in the mirror's one mock data module, keyed by the persona's id, in the same shapes the screens already read. Derived, not duplicated: counts, totals and "3 overdue" come from the data, so every screen agrees. Reuse shared data (catalogs, products) instead of copying it.
- Their texts: data the app localizes follows its languages; names and user content stay as the user wrote them.
- Nothing in the tool (
src/) changes: a persona is the app's.
5. Check and show
npm run typecheck,npm run lint,npm test.- Look at the screens the persona changes, as them:
npm run shot -- "/<page>?persona=<id>", phone and tablet, light and dark. Empty states for a new user, long names that wrap or cut, a paid feature shown or locked, a role's missing actions. - Show the person, short: each persona, who they are, what they change on screen, the link to try them.
- Add an entry to the root
CHANGELOG.md(and bumppackage.json's version): it's a change to the team's mirror.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

