connect-ai-workday
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and act upon metadata from the Workday API, creating an attack surface where a compromised tenant could provide malicious metadata.
- Ingestion points: Metadata enters the agent context through discovery tools:
getSchemas,getTables,getColumns, andgetProcedures(SKILL.md). - Boundary markers: The instructions suggest using square brackets
[]for column quoting, which helps with syntax but does not function as a security boundary against malicious content in metadata. - Capability inventory: The skill allows the agent to perform data retrieval (SELECT), data modification (INSERT, UPDATE), and business process initiation via
executeProcedure(rest.md). - Sanitization: No specific sanitization of external metadata or user input is described beyond standard SQL quoting.
- [SAFE]: The skill contains no executable scripts, hardcoded credentials, or remote network calls. All referenced resources are local markdown files within the skill package, and the instructional content aligns with the stated purpose of database querying.
Audit Metadata