ask-flutter-mechanic

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION] (MEDIUM): The skill directs the agent to execute local shell scripts (./ship-android.sh and ./ship-ios.sh) within the user's project directory. If these scripts are provided by an untrusted source or malicious project, the agent will execute them with the user's full privileges.
  • [EXTERNAL_DOWNLOADS] (LOW): The skill utilizes fvm flutter pub get, bundle install, and pod install. These commands download third-party code from public registries (Pub, RubyGems, CocoaPods), which is standard for Flutter development but carries inherent supply-chain risks.
  • [REMOTE_CODE_EXECUTION] (LOW): The use of build_runner and pod install involves executing external or generated code as part of the build process.
  • [INDIRECT_PROMPT_INJECTION] (LOW): The skill reads and acts upon external data sources which could be manipulated to influence agent behavior.
  • Ingestion points: Reads .fvmrc, gradle-wrapper.properties, and terminal logs for dependency conflict resolution.
  • Boundary markers: None. The agent processes the raw content of these files/outputs directly.
  • Capability inventory: Includes file system modification (rm -rf), package installation, and execution of local shell scripts.
  • Sanitization: No evidence of sanitization or validation of the ingested file contents or terminal output before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:42 PM