html-to-interaction-prompts
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted external data (HTML, CSS, and JavaScript) from live websites or user-provided files to extract interaction logic.
- Ingestion points: The workflow (Step 1) explicitly instructs the agent to read and search through HTML, CSS, and scripts to identify interaction terms like 'ScrollTrigger' or 'animation'.
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when the agent processes the external code, increasing the risk that the agent might follow instructions hidden in the source files.
- Capability inventory: The skill is capable of writing to the file system (creating articles and assets) and executing shell commands (git, ffprobe).
- Sanitization: There is no mention of sanitizing or escaping the external content before the agent analyzes it or interpolates findings into the resulting articles.
- [COMMAND_EXECUTION]: The skill utilizes several shell-based tools to manage the development workflow and verify assets.
- Evidence: The instructions direct the agent to use
git status,git diff --check,git add -f, andffprobefor asset verification and source control management within the article workspace.
Audit Metadata