kelly-wechat-content
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and supporting scripts follow security best practices by implementing least privilege access, scoping operations to a specific folder, and disclaiming any automated publishing capabilities. All write operations are directed through a supervised ChangeRequest process.
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from external editorial sources.
- Ingestion points: Data enters via the
sourcesbase (fields:summary,search-text) andarticlesbase (fields:draft,outline). - Boundary markers: Instructions explicitly require the agent to submit proposals as ChangeRequests for human verification, preventing autonomous execution of embedded instructions.
- Capability inventory: Capabilities are restricted to structured record management (list, get, count, groupBy) within the Busabase workspace; no arbitrary shell execution or external network exfiltration tools are used for processing user data.
- Sanitization: The application frontend implements HTML escaping and validates resource ownership metadata (
appId,resourceKey,schemaVersion) before record processing.
Audit Metadata