zeabur-service-exec
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to run shell commands (e.g.,
ls,cat,env) and database clients (e.g.,mongosh,psql,redis-cli) inside remote containers via the Zeabur CLI. This capability is the primary and intended purpose of the skill for environment management and debugging.\n- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to usenpx zeabur@latest, which downloads the Zeabur CLI from the official NPM registry. As this tool belongs to the skill's author (zeabur), it is considered a safe and legitimate resource.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied service IDs and command strings, creating an attack surface where a malicious user could attempt to manipulate the shell command construction.\n - Ingestion points: The
<service-id>and<command>parameters in the shell command templates are derived from user input.\n - Boundary markers: The skill utilizes the standard
--shell separator to distinguish between CLI flags and the command executed within the container.\n - Capability inventory: The skill provides shell access to targeted containers, which includes reading configuration files and inspecting environment variables.\n
- Sanitization: No explicit validation or filtering of the command string is performed by the skill itself, relying on the underlying agent's intent recognition.
Audit Metadata