leave-management
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted user input to derive field lists and generate technical artifacts, which represents a potential surface for indirect prompt injection attacks.
- Ingestion points: The skill actively captures user-supplied descriptions of existing processes ("our process is..." or "it is in a sheet") and answers to scope and rule questions in Step 1 and Step 2 to build its internal context.
- Boundary markers: There are no technical delimiters (like XML tags or specific tokens) defined to encapsulate untrusted user input within the prompts, though the instructions provide natural language guidance to keep prose outside of data outputs.
- Capability inventory: The skill is capable of generating machine-readable CSV, SQL DDL, and JSON Schema files, and it can route tasks to the
@notion-manual-importagent for workspace integration. - Sanitization: The skill contains explicit 'Security & Safety Notes' that mandate the use of placeholders, the masking of bank details, and the exclusion of real medical or personal identifiers from generated templates.
Audit Metadata