skills/owainlewis/blueprint/html-doc/Gen Agent Trust Hub

html-doc

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the subprocess.run method to execute pandoc and node (for the Mermaid CLI) to perform document transformation. These calls are constructed using argument lists rather than shell strings, preventing typical command injection vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown content and implements a rigorous sanitization pipeline to mitigate injection risks.\n
  • Ingestion points: Markdown source is read from local files or standard input within scripts/html_doc.py.\n
  • Boundary markers: Input is parsed into a structured abstract syntax tree (AST) via Pandoc, and custom transformations are applied to enforce document boundaries.\n
  • Capability inventory: The skill can execute subprocesses for rendering and perform file system operations to manage candidate documents.\n
  • Sanitization: Raw HTML blocks and inlines are escaped, link schemes are restricted to a safe subset (http, https, mailto), and generated SVG diagrams are validated against an allowlist.\n- [DATA_EXFILTRATION]: Multiple network isolation strategies are implemented to prevent data exfiltration. The Puppeteer rendering engine is configured with host-resolver and proxy rules to block all external network access, and the final HTML output is scanned to ensure it contains no external runtime assets or scripts.\n- [DYNAMIC_EXECUTION]: The skill renders Mermaid diagrams using a headless browser. This risk is managed by pre-rendering checks that reject source code containing resource-bearing syntax (like URLs or imports) and post-rendering validation that restricts the SVG output to a safe set of elements and attributes.\n- [EXTERNAL_DOWNLOADS]: The skill instructions include running npm ci to install required dependencies. These dependencies are pinned to specific versions in the package-lock.json file and are sourced from a standard public registry.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 05:01 AM