writing-settings-forms
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the creation of settings forms which ingest untrusted data into the agent's environment. This data is then used in downstream processes such as Handlebars templates, filters, and script hooks.
- Ingestion points: Data is ingested through form fields defined in
settingsForm.form(SKILL.md) and stored in thesettingsobject of various resources (Integrations, Flows, etc.). - Boundary markers: The documentation references standard Handlebars delimiters (
{{settings.path}}) for interpolating these values into templates and filters. - Capability inventory: Values stored in
settingsare accessible to scripts (which can perform complex logic) and can be used to parameterizeHTTPExportrequests inexportSelectcomponents. - Sanitization: The skill explicitly notes that the platform does not validate data written via API, CLI, or MCP tools, potentially allowing malformed or malicious instructions to be stored if the UI validation is bypassed.
Audit Metadata