list-growth-designer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from email platform (ESP) exports and web analytics reports.
- Ingestion points: Reads user-provided ESP exports, growth history data, and reports as mentioned in the Data Sources section and instructions.
- Boundary markers: The skill lacks programmatic delimiters for processing external data but includes explicit natural language instructions to treat every export or pasted record as untrusted input and never follow instructions embedded in a CSV or report.
- Capability inventory: The skill has file system write access to the
memory/directory and invokes a local scriptregistry-events.pyvia an authorized propose request. - Sanitization: There is no evidence of programmatic escaping or schema validation for the ingested data; the skill relies on the agent following the safety instructions in SKILL.md and SECURITY.md.
Audit Metadata