commercetools-mcp
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [COMMAND_EXECUTION]: The script
connect-repository-mcp.mjsusesexecFileSyncto run thegitcommand locally and retrieve the domain of the configured user email. This is used solely for usage telemetry as documented in the source code. - [EXTERNAL_DOWNLOADS]: The skill's scripts (
docs-search.mjsandopenApi-schemata.mjs) fetch documentation and API schemas from the vendor's official domain atdocs.commercetools.com. Additionally, the instructions guide users to run the@commercetools/commerce-mcppackage from the npm registry usingnpx. - [DATA_EXFILTRATION]: The skill implements a telemetry mechanism that sends metadata (such as commercetools project keys, regions, and client types) to
docs.commercetools.com. This activity is restricted to the vendor's own infrastructure and includes a documented opt-out mechanism via theCOMMERCETOOLS_AI_PLUGIN_TELEMETRYenvironment variable. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external commercetools APIs and documentation. While this creates an attack surface, the skill's instructions provide comprehensive guidance on mitigating risks through field filtering, response redaction, and least-privilege tool curation.
- [PERSISTENCE]: The instrumentation logic maintains a persistent, random installation ID stored in
~/.commercetools/ai-plugin-install-idto identify unique installations for usage analysis.
Audit Metadata