Attachments Cleanup Assistant

Remove unused attachments
Check pages for not displayed attachments
Never lose track of your files!
Rovo agent at work
Advanced view ... reclaim some storage!

Inspiration

As Confluence spaces grow, pages accumulate attachments that are no longer needed - outdated images, unused files, and remnants of old templates. These increase storage costs, slow down spaces, and make content harder to maintain.

Deleting attachments is risky. It’s often unclear whether a file is still referenced (in the page body, as a header or cover image), reused elsewhere, or protected by conventions or labels. Teams avoid cleanup because a single wrong deletion can silently break content.

We wanted a safe, explainable, and on-platform way to help teams understand which attachments are truly unused - and why.

What it does

Attachments Cleanup Assistant helps Confluence teams identify and safely clean up unused attachments.

Scans Confluence pages to detect:
- attachments referenced in the page body (including filename-based hints)
- header and cover image usage
- attachments protected by labels or defined rules
Classifies attachments as:
- Used
- Unused
- Protected / requires review
Uses Rovo agent actions to explain why an attachment is considered safe, risky, or protected
Lets users talk to their data:
- “Why is this attachment unsafe to delete?”
- “What would break if I remove this file?”
Provides safe cleanup flows:
- single-attachment deletion with explicit confirmation
- bulk cleanup suggestions that are never destructive by default
Fully permission-aware:
- runs with the initiating user’s permissions
- never exposes content the user cannot access

How we built it

Platform

Atlassian Forge app for Confluence Cloud
Custom UI (React) with Forge Bridge and Atlassian design primitives
Native Rovo agent with action modules for analysis and explanation

Backend

Page and attachment scanning via Confluence REST APIs
Deterministic usage detection based on multiple signals:
- in-body references
- header and cover image usage
- high-confidence filename and metadata hints
Label-based protection rules and conservative risk heuristics
Permission-aware execution using user-scoped access
Structured logging and storage-backed configuration

Frontend

Clear visual indicators for attachment status:
- Used - ✅
- Protected - 🔒
- Unused - ❌
Review-first UX:
- visible evidence and “why” explanations
- explicit confirmations before any destructive action
Resilient UI for large spaces and partial data, with accessible components

Challenges we ran into

Hidden usage detection
Cross-page references are difficult to detect reliably. We prioritized high-signal checks and designed the system to be conservative, encouraging review over automation.
Safety over automation
Full auto-delete is dangerous. The system is designed to explain first and act only with explicit user intent.
Permissions and trust
Actions must respect user visibility. We carefully distinguish re-authentication issues from genuine access restrictions to avoid misleading users.
Scale
Large spaces can contain thousands of attachments. We optimized scans, limited payload sizes, and kept the UI responsive.