experience-cms-content-generate
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes authorized Salesforce MCP tools (create_cms_content, update_cms_content_variant, publish_content) for all content operations, ensuring interactions are confined to official platform services.
- [SAFE]: A two-layer validation strategy is implemented for manual workspace IDs, verifying both the format (0Zu/9Pu prefixes) and existence via server-side checks before use in write operations.
- [SAFE]: Content generation and editing are strictly server-side processes; the agent only provides user goals verbatim as prompts, preventing local manipulation of sensitive schema structures.
- [SAFE]: Local filesystem operations are restricted to scanning for UIBundles and persisting non-sensitive channel IDs within the project's own directory structure.
- [INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted data from user messages (SKILL.md Step 1), content variant bodies (references/edit-publish-workflow.md), and sibling skill outputs (references/delegation-protocol.md). Boundary markers are not used to delimit these inputs. Capability inventory includes CMS content writes, local configuration file writes (references/workspace-resolution.md), and cross-skill execution. Sanitization is not performed by the agent, as all resulting content is validated and generated by the backend Vibes MC content pipeline.
Audit Metadata