platform-salesforce-connect-adapter-generate
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill generates Apex classes that fetch and process JSON data from external REST endpoints. This creates an attack surface for indirect prompt injection, as malicious or compromised external services could return data structured to exploit the agent's logic or the processing flow within Salesforce.
- Ingestion points: External REST API response bodies (res.getBody()) processed in the generated DataSource.Connection Apex classes.
- Boundary markers: Absent in the generated Apex templates for data ingestion.
- Capability inventory: Http.send() for network requests in Apex, Database.query() in verification controllers, and sf project deploy start for metadata deployment.
- Sanitization: Uses standard JSON.deserializeUntyped() and EncodingUtil.urlEncode() for parameters.
- [COMMAND_EXECUTION]: The skill provides instructions and examples for using the Salesforce CLI (sf project deploy start) to deploy the generated Apex code and metadata. This is a standard developer workflow for the Salesforce platform and is scoped to deploying the skill's own generated artifacts to a targeted org.
Audit Metadata