atlassian-mcp

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Documentation recommends installing official and community-sourced MCP servers using package managers such as npm, pip, and uv. It specifically references the official Anthropic repository for the Atlassian MCP server.
  • [COMMAND_EXECUTION]: Provides instructions for configuring and running MCP servers via command-line tools like npx and uvx within the Claude Desktop environment.
  • [PROMPT_INJECTION]: The triage and documentation synchronization workflows implemented in 'references/common-workflows.md' ingest untrusted content from Jira and Confluence.
  • Ingestion points: Reads issue summaries and descriptions via 'jira_get_issue' and page content via 'confluence_get_page'.
  • Boundary markers: Implementation examples do not include explicit delimiters or instructions for the agent to ignore embedded instructions in the ingested data.
  • Capability inventory: The skill has the capability to write back to the system via 'jira_update_issue', 'jira_add_comment', 'confluence_create_page', and 'confluence_update_page'.
  • Sanitization: While an 'escapeHtml' function is provided for Confluence storage format, the logic determining triage categories and priority is based on raw text patterns without structural sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 11:42 AM