The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the mcp__serena__execute_shell_command tool to run a local script located at ./scripts/validate-gate-3a.sh. Executing local scripts whose contents are not defined within the skill itself presents a risk of arbitrary command execution on the host system.
[DATA_EXFILTRATION]: The skill contains a hardcoded absolute file path /Users/nick/Desktop/claude-mobile-expo within the cwd parameter of a shell command. This exposes the local username and directory structure of the developer environment, which is a form of information disclosure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via its visual verification workflow.
Ingestion points: The agent takes screenshots of the mobile application UI using expo-mcp and performs visual analysis on the content.
Boundary markers: None. The skill does not instruct the agent to ignore text or instructions rendered within the screenshots.
Capability inventory: The agent has the ability to execute shell commands (mcp__serena__execute_shell_command), read files (mcp__serena__read_file), and write to memory (mcp__serena__write_memory).
Sanitization: None. If the application under test displays attacker-controlled content (e.g., a message or a profile name), that content could contain malicious instructions that the AI might interpret as commands while performing visual verification.

claude-mobile-validation-gate