atlassian-mcp

Warn

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install and run third-party, community-maintained MCP servers such as @sooperset/mcp-atlassian and atlassian-mcp. These packages originate from unverified sources outside of recognized trusted vendors.\n- [COMMAND_EXECUTION]: Configuration examples provided for Claude Desktop involve the use of npx and uvx to download and execute remote code during the initialization of the MCP environment.\n- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection through its automation workflows.\n
  • Ingestion points: The triage and documentation sync workflows in references/common-workflows.md read untrusted content from Jira issue fields (summary, description) and Confluence page content.\n
  • Boundary markers: The implementation lacks delimiters or instructions to ignore potential commands embedded within the processed external data.\n
  • Capability inventory: The skill has significant write capabilities across the Atlassian suite, including creating and updating issues and pages, which could be abused if malicious instructions are processed.\n
  • Sanitization: There is no sanitization of input data for logic control; sanitization is only applied to output formatting (e.g., escapeHtml).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 3, 2026, 09:32 AM
Security Audit — agent-trust-hub — atlassian-mcp