The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of the Maestro CLI through the command curl -Ls "https://get.maestro.mobile.dev" | bash. This pattern executes remote code directly in the shell without any verification or hash-checking.
[COMMAND_EXECUTION]: The instructions direct the agent to expand its own permissions by modifying builder.config.json. The requested configuration uses wildcards for powerful system commands (e.g., curl *, chmod *, rm *, kill *), which significantly broadens the agent's attack surface and allows for arbitrary system interaction.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes installation scripts from the external domain mobile.dev during the environment setup phase.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted data from the local repository.
Ingestion points: Reads and processes YAML flow files in the maestro/ directory and application code in ios-app/ and android-app/.
Boundary markers: None; there are no instructions or delimiters to prevent the agent from following instructions embedded within the test flows or source code.
Capability inventory: Extremely high; the requested ACL expansion provides the agent with network access, file system manipulation, and process control.
Sanitization: None; the skill does not perform any validation of the content read from the project files.

mobile-testing