docs-features
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to parse and transform markdown content that may originate from untrusted sources, particularly when using custom comment-based components like
<!-- ::start:tabs -->. - Ingestion points: Untrusted data enters the agent context through the
sourcestring passed toparseMarkdowninSKILL.md. - Boundary markers: The skill uses specific comment-based delimiters (
<!-- ::start:name -->and<!-- ::end:name -->) to identify and scope custom component blocks. - Capability inventory: The skill generates HTML and metadata based on parsed content, including
md-tab-panelandmd-framework-panelelements with associated data attributes. - Sanitization: The documentation states that component properties are serialized to JSON and escaped as HTML attributes during rendering, providing a layer of protection against direct script injection.
Audit Metadata