clawvard-agent-eval
Pass
Audited by Gen Agent Trust Hub on May 8, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill incorporates a bash script that automatically checks for updates and performs a global installation of the skill using
npxfrom the vendor's repository. - [COMMAND_EXECUTION]: Shell commands are utilized within the pre-flight scripts to perform version validation and package management, specifically using
curl,stat, andnpx. - [EXTERNAL_DOWNLOADS]: The skill performs network requests to
https://raw.githubusercontent.comfor version checks and communicates withhttps://clawvard.schoolto facilitate the exam process. - [DATA_EXFILTRATION]: The skill transmits agent metadata (model ID, performance traces) and authentication tokens to external endpoints at
https://clawvard.school. - [PROMPT_INJECTION]: The skill processes structured and unstructured data from an external API, which represents a surface for indirect prompt injection.
- Ingestion points: API responses retrieved from
https://clawvard.school/api/exam/*. - Boundary markers: None explicitly defined in the instruction set to separate API content from agent instructions.
- Capability inventory: Ability to execute shell commands via the update script and perform outbound network operations.
- Sanitization: The skill does not specify validation or sanitization procedures for data received from the external evaluation service.
Audit Metadata