1. Architecture

• Forge hosting: all app functions execute on Atlassian Forge.
• No external backends: we do not operate separate servers that receive Jira tenant data.
• Forge LLMs: AI requests use Atlassian-hosted models; issue text in prompts stays on-platform.
• Async workers: long-running AI tasks run in Forge async event consumers to avoid trigger timeouts.

2. Least Privilege

Requested scopes:

• read:jira-work — read issues, comments, and available transitions
• write:jira-work — add comments, update descriptions, perform transitions
• storage:app — persist enabled space IDs from the admin page

The app processes issues only in spaces enabled by a Jira administrator.

3. Loop Prevention & Site Safety

• Manifest trigger filters limit invocations to Magic Column status transitions.
• ignoreSelf: true and actor checks prevent the app from reacting to its own updates.
• Product triggers never throw errors back to Jira—failures are logged only.

4. Secure Development

• Dependency vulnerability checks (npm audit) before release.
• No secrets or credentials in source repositories.
• Logging avoids tokens, passwords, and unnecessary personal data.
• TLS for all platform communications.

5. Vulnerability Reporting

Report security issues to support@lndtech.eu with reproduction steps. We triage promptly and coordinate with Atlassian via ECOHELP when required.

References:

• Atlassian Marketplace security requirements
• App approval guidelines