platform-custom-application-generate
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to query Salesforce data and generate metadata, creating an attack surface for indirect prompt injection. 1. Ingestion points: Application labels, API names, and profile names provided by the user in SKILL.md. 2. Boundary markers: The skill requires a manual confirmation step where the agent must restate the plan and wait for user approval before executing any write operations. 3. Capability inventory: The skill can perform SOQL and Tooling API queries and execute metadata operations (POST/PUT/DELETE) to create apps, FlexiPages, and modify profile-based access permissions via SetupEntityAccess. 4. Sanitization: No explicit sanitization logic is provided in the skill instructions, although the instructions note that target API endpoints validate metadata names.
Audit Metadata