atlassian-mcp

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Outsider-authored free text from Atlassian Jira/Confluence (e.g., issue summaries/descriptions/comments and Confluence page body.storage) is fetched at runtime via MCP tool calls like jira_get_issue, jira_search, confluence_get_page, and confluence_search, then parsed/embedded into the agent’s LLM context for downstream generation (e.g., triage, release notes, meeting-note extraction).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill includes runtime server startup commands that fetch and execute remote npm packages (e.g., "npx -y @sooperset/mcp-atlassian" and "npx @anthropic/mcp-atlassian" shown in the MCP server configuration/setup), which means remote code from the npm registry would be retrieved and executed as a required runtime dependency.