html-spec-planning
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided content to generate HTML documents, which establishes a potential surface for indirect instruction injection. To mitigate this, the skill strictly mandates safe DOM construction (using
textContentandcreateElementinstead ofinnerHTML) to prevent XSS within the generated files. - Ingestion points: Processes user requests for specifications, implementation plans, and RFCs.
- Boundary markers: Output is isolated to standalone files, which helps prevent leakage into the primary agent context.
- Capability inventory: Writes HTML files to the filesystem and includes a mechanism for platform integration via
submitToClaude. - Sanitization: Mandatory avoidance of
innerHTMLfor dynamic content prevents execution of injected scripts within the artifact. - [EXTERNAL_DOWNLOADS]: The generated HTML artifacts reference external font resources from Google Fonts, which is a standard and safe practice for document styling.
Audit Metadata