aguasto-integration-testing
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references external resources hosted on
www.aguasto.com, including an OpenAPI contract (/openapi.json), a sandbox guide (/sandbox), and a sandbox API endpoint (/api/v1/sandbox). These are used to define the schema and environment for integration testing. - [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests data from an external API, creating a surface for potential indirect prompt injection.
- Ingestion points: API responses from
https://www.aguasto.com/api/v1/sandbox(documented inSKILL.md). - Boundary markers: The instructions include a defensive guideline to 'Read structured fields rather than matching prose' when processing responses.
- Capability inventory: Mentions the
aguasto-research fixturesCLI command for local verification. - Sanitization: No specific sanitization or filtering logic is defined for the external API content within the skill instructions.
- [COMMAND_EXECUTION]: The skill mentions the usage of a CLI tool,
aguasto-research, to perform local-only SDK checks and exercise bundled fixtures. This is presented as a development and testing utility.
Audit Metadata