react-native-best-practices

Warn

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill recommends installing the Flashlight performance tool using a 'curl pipe to bash' pattern (curl https://get.flashlight.dev | bash), which allows for the execution of unverified remote scripts.
  • [PROMPT_INJECTION]: The skill exhibits metadata poisoning; the internal metadata attributes the content to 'Callstack', while the provided author context identifies 'pyinx', which is misleading regarding the origin and trustworthiness of the provided instructions.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface. Ingestion points: Reads local project configuration files like package.json and .gradle via search commands. Boundary markers: No delimiters or ignore-instructions warnings are present. Capability inventory: Executes build tools (gradlew, xcodebuild), package managers (npm, npx), and network requests (curl). Sanitization: No escaping or validation of project-derived content is performed.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute high-privilege shell commands including gradlew and xcodebuild to interact with native build systems.
  • [EXTERNAL_DOWNLOADS]: The skill fetches tools and configurations from third-party services like Spotify's Ruler and Emerge Tools for app size analysis and performance monitoring.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 28, 2026, 01:45 PM