
from fazm219
Query a locally-extracted browser profile (emails, accounts, tools, contacts, addresses, payments) so agents can answer personal-context questions without sendi
This skill exposes a local, read-only browser profile database containing autofill data, saved logins, history, bookmarks, and extracted contact/payment info. It lets agents answer user-specific questions ("What's my email?", "Which accounts do I have?", "What payment card did I use?") by querying a local store — no data leaves the machine. The skill provides a simple query API (query_browser_profile) that returns ranked, self-scoring results for identity, contacts, accounts, tools, addresses, and payment details.
Use this skill whenever the agent needs personal context to give accurate, actionable answers about the user's own information. Typical triggers: recovering contact details, listing linked accounts, summarising tools and services the user uses, finding a saved address or payment card, or producing a compact user profile. If queries return empty, run the provided extractor to refresh local data.
Best used with agents that can call local tools or native integrations (desktop/macOS agents, assistant frameworks with filesystem access). Compatible agent types: local-desktop agents, Claude/Copilot-style assistants with native connectors, or any agent runner that supports invoking a small local query API.
AI Browser Profile is a skill that lets agents query a locally-extracted browser profile (emails, accounts, contacts, addresses, payments) stored in a local SQLite database. No scripts were included — the skill is purely instructional, relying on a built-in tool (query_browser_profile) and a native Swift extractor. Documentation is clear with good usage examples, but the skill has a macOS/Swift dependency for the extraction step and lacks error-handling guidance.
No scripts to run or evaluate. Skill is entirely instructional, delegating to a built-in tool. The local-only data storage is good for privacy. The main concern is the breadth of sensitive data accessible without explicit access-control or redaction guidance, though this is by design for personal-context skills.