build-your-users-mind — Public, provider-neutral pointer

build-your-users-mind — Public, provider-neutral pointer

This skill is a thin pointer to the public module ellmos-ai/build-your-users-mind. The module contains the complete recipe, templates, schemas, scripts, tests and source-adapter documentation. This catalog does not duplicate that code.

What the module does

With the operator's explicit authorization, the module helps an agent:

1. extract genuine user-authored turns from the operator's own interaction logs;
2. redact sensitive material before persistent storage;
3. reduce and classify evidence about recurring preferences and decisions;
4. create a local preference model with confidence levels and provenance;
5. bind a short pointer into the selected agent runtime; and
6. calibrate predictions against real later feedback.

The public module is a recipe for any user and any supported agent runtime. It does not contain a model of a specific person.

Safety and privacy boundary

• Operator authorization is required before reading interaction logs.
• Personal profiles, raw logs, evidence corpora and local paths remain private.
• Predictions are uncertain hypotheses, not mind-reading, diagnosis or statements made by the user.
• A preference prediction never expands the agent's authority.
• External, irreversible, safety-critical, legal, medical, employment, financial or similarly high-impact actions require explicit confirmation.
• Agent-generated predictions must never become primary evidence about the user.

Installation

git clone https://github.com/ellmos-ai/build-your-users-mind.git <clone-path>