fcode-forms
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the creation and use of web forms that ingest untrusted user input, which represents a surface for indirect prompt injection attacks against backend processes.\n
- Ingestion points: Input parameters collected through forms embedded via the SDK or React components.\n
- Boundary markers: Instructions state that result messages are rendered as markdown and that raw HTML is never rendered or executed.\n
- Capability inventory: Backend processes triggered by form submissions can execute JavaScript or Python logic and interact with file storage.\n
- Sanitization: The platform automatically strips raw HTML from messages to prevent script injection.\n- [EXTERNAL_DOWNLOADS]: The skill instructs the user to include an external JavaScript SDK from the platform's domain.\n
- Evidence:
https://code.factorialhr.com/sdk/forms.jsis required for form initialization in HTML embeds.\n- [REMOTE_CODE_EXECUTION]: The skill facilitates the execution of backend code (Factorial Code processes) triggered by frontend form submissions.\n - Evidence: Mentions of
preRenderProcessand form-triggered process executions that run server-side logic.\n- [COMMAND_EXECUTION]: The skill describes how backend processes handle logic, which may include calling other processes or managing data, though no direct shell command execution is exposed to the frontend.\n- [NO_CODE]: This skill consists of documentation and configuration instructions rather than executable script files directly bundled in the skill package.
Audit Metadata