soulscrape
distill the essence of any human, for reference, imitation, or fun.
soulscrape is an agent skill that turns the evidence you're authorized to use into a dated working model of a person: how they decide, write, argue, and change their mind, with every claim wired to its source and every contradiction left in. use it as a collaboration guide, a personal operating manual, or the bootstrap for an assistant that works like they do.
the model describes patterns in the supplied evidence. it does not establish a complete identity, a diagnosis, consent, or the right to impersonate or act for someone. reference is free; imitation needs the subject's own sign-off.
free to use, with your own agent
Your agent performs the research and synthesis with your model, tools, and authorized sources. The complete Skill runs in your agent environment without a Soulscrape account. Your agent or research providers may charge for their services; those costs are separate.
A free Hraness account is needed only to publish and manage public indexes. Reading public pages and read APIs is free without sign-in. No Soulscrape subscription, Credits, or payment card is required. When you choose to publish, Hraness receives the reviewed public packet and the account/device information needed to manage it. Private sources and working documents stay in your chosen agent environment, subject to that environment's data practices.
install and build your first model
Use Bun 1.3.14 or newer and a compatible agent, such as Codex or Claude Code. Review the skill, then install its pinned release:
bunx skills add hraness/soulscrape#v0.4.1 --skill soulscrape
Installation is inert: it does not inspect personal data or start a modeling run. Start a new agent session, supply a small corpus you are authorized to use, and make the purpose and audience explicit. For example, provide your own decision log and weekly notes, then ask:
Use $soulscrape to build a private self-model from the decision log and
weekly notes I provided. Write a standalone Markdown personal operating
manual for my own use. State the source date range, tie claims to evidence,
and show counterevidence and uncertainty. Do not use public web research
or prepare an assistant charter.
The agent maps those sources, resolves material gaps in scope, and writes a new Markdown document by default. Review the claims against their sources and correct the model before reusing it. Your agent environment determines how supplied material is processed; the skill does not provide its own private hosting or storage service.
see the artifact first
The document follows the evidence rather than a personality template. Its core sections explain the observed patterns, their practical implications, and their limits:
# <Name>: a dated working model
> Status: Partial, source-bounded, dated, and revisable. The real person's
> current words, choices, and corrections outrank this document.
## Executive model
The patterns that best explain the subject's demonstrated choices,
with source references and counterevidence.
## Practical operating manual
How to bring context, disagree, decide, draft, verify, and close loops.
## Tensions, limits, and revision hooks
Where the evidence conflicts, where the model predicts poorly, and what
new evidence should change it.
## What not to infer
Sensitive, unsupported, stale, or out-of-scope conclusions.
soulscrape separates facts, stated beliefs, revealed patterns, and speculation. a pattern supported by engineering decisions can still be untested outside work. contradictions, historical change, and alternative explanations remain part of the model.
how a person becomes a model
- map the authorized corpus. record authorship, source type, date range, audience, sampling limits, and likely blind spots before interpreting it.
- build an evidence ledger. connect each claim to its supporting material. repeated decisions and costly behavior usually carry more signal than polished self-description, while each source keeps its context.
- calibrate the interpretation. assess support and scope separately. preserve counterevidence, uncertainty, and plausible alternative readings.
- write usable guidance. explain what collaborators or an authorized assistant can do with the model, which decisions remain with the person, and when to revise it.
the asking protocol defines the question packet and stop conditions. public research is off by default; when the user enables it, findings retain their URL, access date, and supporting passage, and attribution requires an identity anchor.
the same distillation, served publicly. ask the skill to assemble a public person index — a thinker, a builder, anyone — and it produces a soulscrape.person-index.v1 packet: a cited claims ledger, a timeline, themes, works, appearances, typed relations to other entities, and explicit open questions. you review the complete packet, then publish it with a free Hraness account at soulscrape.com/<username>/<handle> — see the eugene tssui index for a live one — where every claim links to its source and the packet is served as HTML, Markdown, and JSON. Agents can inspect the complete public-read and authenticated publishing contract in the OpenAPI 3.1 document.
