Feedback: Enable create/update access for AI-assisted workflows

Title: Feedback: Enable create/update access for AI-assisted workflows

Summary

Users expect the AI assistant to directly create and update objects (blogs, references, citations) within their Capacities workspace to realize end-to-end automation, reduce manual copy/paste, and maintain consistency across exports (Markdown/HTML) and downstream publishing.

Requested capabilities

Write permission to create new objects of relevant types (e.g., BlogPost, Reference, Weblink, Citation) when guided by a user-provided input payload.

Update permissions to modify existing objects (e.g., attach references, add or adjust fields, add versioning/change logs) as part of the AI-assisted workflow.

Read access should remain available as it is today, but write access is essential for productive, low-friction workflows.

Permission model: allow granting the AI agent access within a chat thread context, with explicit user consent, and a scoped, revocable token or session-based permission.

Operational expectations

Changes should be fully auditable: include a ChangeLog entry or version tag on BlogPost objects whenever content or references are updated.

Non-destructive defaults: if the user doesn’t want to modify an existing object, the AI should present a clear, explicit option to create a parallel draft (e.g., “Propose new BlogPost v2”).

Safety and governance: provide confirmation prompts before making edits, and surface potential conflicts (duplicate references, mismatched reference-ids) for user review.

Proposed UX patterns

In-chat prompts to authorize actions: “Grant temporary write access to workspace for this session?” with a scoped duration.

Preview before apply: show a delta of changes (new objects and updates) and require user confirmation to apply.

Clear provenance: always link to the source of changes (input payload, added references, Changes) in the blog post’s ChangeLog.

Impact on workflows

This capability would streamline the end-to-end process you described: drafting in Capacities, linking citations, rendering for WordPress, and building a reusable citation library within the same workspace.

It would reduce manual handoffs and minimize drift between draft and published formats.

Would you like me to package this as a formal feature request or a quick-start spec you can share with the Capacities product team? If you want, I can draft:

A short, implementation-friendly spec (API/UX surface, object types, and permissions)

A sample risk/mitigation table (security, data integrity, rollback scenarios)

A minimal user-flow diagram description you can attach to the ticket