configure-site

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local shell commands for Git repository management, including git init, git add, and git commit, to scaffold documentation sites. It also employs sed for the batch substitution of cross-space link sentinels with real Space IDs in markdown files after programmatic creation in GitBook.\n- [EXTERNAL_DOWNLOADS]: The skill supports the registration of OpenAPI specifications from remote URLs for auto-generating API documentation. It also makes standard management requests to the GitBook API at https://api.gitbook.com/v1.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by ingesting external content from repositories, folders, or documentation URLs to design site architecture.\n
  • Ingestion points: Reads markdown content and metadata from user-specified repositories or website URLs to designed site structure and populate pages (as described in SKILL.md and references/migration-from-other-platforms.md).\n
  • Boundary markers: Absent. While the skill instructs the agent to verify content sources, it does not specify the use of delimiters when processing external content into prompts.\n
  • Capability inventory: Access to local filesystem for monorepo scaffolding and network access to the GitBook API for site configuration.\n
  • Sanitization: The skill focus is on structural conversion and does not execute the ingested content as code during processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 07:34 PM
Security Audit — agent-trust-hub — configure-site