render
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from local files and outputs of other skills to generate interactive HTML documents.
- Ingestion points: The workflow in SKILL.md resolves content from specified file paths or the results of wrapped skills to populate report templates.
- Boundary markers: There are no explicit instructions for using delimiters or boundary markers to prevent the agent from following instructions embedded within the source documents or skill outputs.
- Capability inventory: The agent uses Write and Artifact tools and is instructed in Step 6 to act on verdicts and comments read back from the generated pages.
- Sanitization: While a script to escape JSON characters in the state block is provided, there is no requirement for sanitizing the primary text content inserted into the HTML templates.
- [DYNAMIC_EXECUTION]: The skill generates HTML files containing embedded JavaScript logic for UI functionality and platform interaction.
- Evidence: Templates in references/feedback-loop.md and references/page-kit.md include scripts that manipulate the DOM, manage localStorage, and call platform-specific APIs like claude.use("artifact") for publishing content.
Audit Metadata