lexoffice
Warn
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: MEDIUMMETADATA_POISONINGEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's frontmatter and overview contain misleading information. The description claims support for managing 'Leads' and 'Pipelines,' which are CRM features not typical of the Lexoffice accounting platform. Furthermore, the 'Lexoffice Overview' section includes an extensive list of generic nouns (e.g., 'Cookie', 'Decision', 'Zone', 'Story') that do not correspond to Lexoffice API endpoints, suggesting deceptive or poorly generated metadata.
- [EXTERNAL_DOWNLOADS]: The instructions require the installation of the
@membranehq/clipackage from the npm registry. This is the vendor's official tool for managing connections and authentication. - [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the Membrane CLI tool (e.g.,
membrane login,membrane action run,membrane request) to interact with external APIs. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Data is ingested from external Lexoffice API responses, including invoice details, contact records, and article descriptions.
- Boundary markers: There are no instructions provided to wrap or delimit external data to prevent the agent from interpreting instructions contained within the records.
- Capability inventory: The skill allows for writing data (creating invoices, updating contacts) and performing arbitrary HTTP requests via the proxy command.
- Sanitization: The skill does not mention any sanitization or validation of the data retrieved from Lexoffice before it is processed by the agent.
Audit Metadata