The Installable Person

The most valuable thing you know isn't written down anywhere. It's the pattern-matching you do in the first ninety seconds of a pitch, the way you rewrite a cold email before sending, the pricing call you make on instinct because you've made it wrong twice before. That judgment has always been trapped in one delivery mechanism: your calendar. Starting now, it can be a plugin.

Software eats things in a predictable order

First it ate documentation. The knowledge that lived in binders and senior engineers' heads moved into READMEs, wikis, Stack Overflow. Static knowledge, but at least it was copyable.

Then it ate workflows. CI pipelines replaced the person who knew the deploy ritual. Zapier replaced the ops hire who moved rows between systems. The procedure became executable, so the procedure stopped needing a human host.

What it never ate was mentorship. Documentation tells you what the API does. A workflow runs the same steps every time. Neither one can look at your specific situation and say "you're solving the wrong problem, here's what I'd do instead." That required a person, which meant it required their time, which meant the best judgment in the world scaled at exactly one conversation per time slot.

Claude Code plugins break that constraint, almost by accident. Plugins have four surfaces: skills, commands, hooks, and MCP servers. Three of those are plumbing. Skills are the interesting one, because a skill is not a script. A skill is a SKILL.md file, YAML frontmatter with a name and a description, plus an optional references/ directory, loaded on demand when the model decides it's relevant. It's prose that changes how the model thinks. Which means it can encode the thing documentation never could: how a specific person reasons.

Judgment as an artifact

Here's what an installable person actually is, mechanically.

You connect your raw material: files, Google Drive, Calendar, call transcripts. installs.me synthesizes that into a persona plugin, a set of Claude Code skills that capture your frameworks, your voice, your priors, and crucially, your published track record so the advice is grounded in things you actually did, not vibes.

Take the demo persona, Lautaro Schiaffino. Argentine founder, sold Sirena to Zenvia for $30M, now runs Darwin AI, angel invests across LatAm, writes at lauta.blog. His persona plugin isn't a chatbot cosplaying an accent. It's skills that trigger on real work: "advise this founder," "draft this in my voice," "would I take this deal." The references/ directory carries the receipts, the frameworks he's published, the postmortems, the way he actually structures a pricing argument. When the model loads the skill, it stops giving you the median internet answer and starts giving you his answer.

That distinction is the whole product. A generic model averages everyone's opinion into something safe and useless. A persona skill is a deliberate bias, and bias is what advice is.

The distribution part is boring, which is the point

Installing a person takes two commands:

/plugin marketplace add https://installs.me/lautaro
/plugin install lautaro@lautaro-installs

Under the hood, that URL serves a marketplace.json, the manifest that tells Claude Code what plugins exist and where their source lives. One sharp edge we hit building this: plugin sources served over a URL must be git-backed. Point a marketplace entry at a relative HTTP path and it fails silently, no error, no plugin. The git-subdir source type is the one that works, verified on Claude Code CLI 2.1.x. The plugin's .claude-plugin/ directory carries the manifest, the skills sit alongside, and from that moment the persona is available in every session, in every repo, on that machine.

Compare that to the old distribution channels for judgment. A book: two years to write, frozen at print, can't see your context. A course: same, with worse production values. An advisory call: sees your context, costs $500 an hour, doesn't scale past the calendar. A persona plugin sees your actual code, your actual draft, your actual deal memo, and responds with a specific person's reasoning, for the marginal cost of tokens.

What changes when you're installable

For the person being installed: your judgment compounds instead of evaporating. Every framework you've written, every call you've taken notes on, becomes training material for a version of you that answers at 3am in someone else's terminal. The advisory work you turn down for lack of time becomes work your plugin takes. You're not replaced by it. You're syndicated.

For the person installing: mentorship stops being gated on access. The founder in Rosario who could never get twenty minutes with an operator who's done a $30M exit can install one. Not a summary of his blog posts. The reasoning engine, pointed at her problem.

And for the honest objection, "it's not really him": correct, and irrelevant. A book isn't really the author either. The question is never whether the artifact equals the person. It's whether the artifact carries enough of the person's judgment to change your decision. A well-built persona skill, grounded in real documents and real track record, clears that bar in a way no static format ever has, because it's the first format that can read your situation before it answers.

Documentation made knowledge copyable. Workflows made procedure executable. Persona plugins make judgment installable. The people who write down how they think, in a form a model can load, are about to be in a lot more rooms than their calendars allow.

Install a person

installs.me turns your files, calendar and calls into a Claude Code plugin that thinks like you. Anyone installs it with two commands:

/plugin marketplace add https://installs.me/lautaro
/plugin install lautaro@lautaro-installs