configure-site

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and instructs the user to connect to the official GitBook MCP server at https://mcp.gitbook.com/mcp. This is a vendor-owned service intended for site management.
  • [COMMAND_EXECUTION]: The skill performs several categories of command execution:
  • Repository Management: Uses gh repo create and glab repo create to initialize remote documentation repositories.
  • Scaffolding: Uses git init, git add, and git commit for local repository setup.
  • Content Processing: Uses sed and find to perform bulk string replacement for cross-space link resolution in markdown files.
  • Package Management: Instructions within tutorials (e.g., references/example-site/guides/tutorials/payment-flows/accept-one-time-payment.md) include npm install and pip install for SDK setup. All command usage is direct and aligned with the skill's primary purpose of documentation site orchestration.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect injection as it processes untrusted user-supplied data (content seeds).
  • Ingestion points: The skill reads user-named repositories, folders, or documentation URLs to design site structures and scaffold content (SKILL.md).
  • Boundary markers: Absent. The skill does not explicitly define delimiters when interpolating source content into its design reasoning.
  • Capability inventory: The skill has significant capabilities including file system writes, shell command execution (via git/gh), and network operations via the GitBook REST API.
  • Sanitization: No specific sanitization or filtering logic is described for the ingestion of source content seeds.
  • [CREDENTIALS_UNSAFE]: The skill manages a sensitive GITBOOK_TOKEN. While it uses environment variables (a safe practice), it provides instructions for users to paste the token directly into the session. However, it includes robust safety instructions: it mandates the use of export, warns the agent never to echo the token back, and explicitly forbids writing the token to any file or committing it to version control.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 12:56 AM