portable-text-conversion

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation provides instructions to install standard development packages including @portabletext/block-tools, @portabletext/markdown, @sanity/schema, and jsdom.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data (HTML and Markdown) from legacy CMSs, APIs, and local files, creating a potential surface for indirect prompt injection if the processed content contains malicious instructions.
  • Ingestion points: Content is passed to htmlToBlocks and markdownToPortableText as shown in rules/html-to-pt.md and rules/markdown-to-pt.md.
  • Boundary markers: The preprocessHtml function demonstrates basic content filtering (e.g., removing script tags).
  • Capability inventory: The provided code examples demonstrate network requests (fetch), file system reads (fs.readFileSync), and data writes to a Sanity dataset via the @sanity/client library.
  • Sanitization: The skill suggests pre-processing HTML to strip potentially dangerous tags like <script> and <style>.
  • [DATA_EXFILTRATION]: The skill includes code snippets for downloading image assets from external URLs using fetch as part of the content migration workflow. This is a functional requirement for its stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:27 AM
Security Audit — agent-trust-hub — portable-text-conversion