Skip to content

MVP_SCOPE.md

Latest commit

 

History

History
114 lines (87 loc) · 4.02 KB

MVP_SCOPE.md

File metadata and controls

114 lines (87 loc) · 4.02 KB
Raw

MVP Scope

This document defines the minimum useful product boundary for OpenChainClaw. It focuses on the public product scope and avoids private planning details.

MVP Goal

The MVP should let a personal user run local assistant tasks and verify the critical action trail.

A usable MVP must answer these questions:

  • What did the assistant read, change, visit, call, or request?
  • Which operations were risky, blocked, approved, or rejected?
  • What files changed, and can they be restored?
  • Do the local audit hash and proof record match?
  • Did raw private data stay out of external records and analytics?

Must Have

Local Task Console

  • create a task;
  • show task status;
  • show a timeline of key operations;
  • show pending approvals;
  • show audit reports;
  • expose rollback actions for supported file changes.

Assistant Runtime

  • execute at least one real local file task;
  • execute at least one real web task;
  • execute at least one real external API task;
  • record purpose summaries for important operations;
  • preserve a clear state model for pending, blocked, failed, approved, rejected, rolled back, and completed states.

File Safety

  • block hidden or sensitive files by default;
  • require explicit authorization for configured directories;
  • create snapshots before mutation;
  • show text diffs when possible;
  • validate rollback inputs;
  • record rollback actions in the audit trail.

Web And API Safety

  • record target URLs and API service names;
  • maintain an allowlist for low-risk targets;
  • pause non-allowlisted or sensitive outbound actions for approval;
  • redact request metadata before audit, analytics, or proof export;
  • mark paid or privileged API calls before execution.

Audit And Proof

  • maintain structured task events;
  • generate deterministic local audit hashes;
  • write local verifiable records;
  • compare local records with proof metadata;
  • keep raw prompts, file contents, diffs, API bodies, tokens, private keys, cookies, and browser credentials out of external proof records.
  • preserve a path for future policy proofs or selective disclosure without requiring public task histories.

Analytics Privacy

  • analytics must remain opt-in;
  • analytics events must use safe metadata only;
  • raw prompts, content, paths, credentials, and identifiers must not be sent.

Should Have

  • guided first-run setup;
  • authorized directory management;
  • website allowlist management;
  • user preferences for default risk behavior;
  • audit search by task, risk level, operation type, path, URL, or API service;
  • report export to Markdown or JSON;
  • failure and retry states for proof submission;
  • wallet connection as an optional proof or approval enhancement.

Could Have Later

  • richer browser automation adapters;
  • mobile messaging status notifications;
  • limited mobile approval summaries;
  • optional companion app;
  • additional proof backends;
  • privacy-preserving proof or reputation backends;
  • plugin or connector SDK;
  • team workflows;
  • enterprise compliance reports.

Not In The MVP

  • automatic execution of privileged actions without local review;
  • raw private data on-chain or in externally verifiable records;
  • reading wallet private keys or seed phrases;
  • treating mobile messaging channels as trusted local consoles;
  • group chat control by default;
  • full enterprise access control;
  • multi-agent governance;
  • public agent identity, feedback, delegation, or reputation graphs that link a user's activity across contexts by default;
  • public publishing of private task histories;
  • broad background automation before approval, rollback, and audit behavior are stable.

Release Boundary

The MVP should not be considered ready until:

  • high-risk operations pause before execution;
  • blocked operations fail closed;
  • file mutation has a recovery path;
  • audit events are structured and searchable enough for user review;
  • proof records are verifiable without exposing raw private content;
  • analytics privacy tests cover the current event catalog;
  • documentation explains safety defaults, user flows, and known limitations.