patrol
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from the local development environment.
- Ingestion points: Reads
pubspec.yaml, Dart source files (lib/keys.dart, feature modules), test files (patrol_test/*.dart), and device logs. - Boundary markers: The instructions do not specify the use of delimiters or ignore-instructions for the data read from the filesystem.
- Capability inventory: The skill has the ability to write files (creating/modifying keys and modules) and execute commands via the Patrol CLI (using tools like
patrol-run). - Sanitization: There is no evidence of input validation or sanitization of the content read from project files before it is processed by the agent.
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to use various tools (
patrol-run,patrol-screenshot,patrol-native-tree,patrol-status,patrol-quit) which execute shell-level commands to drive mobile automation and interact with emulators/devices. - [EXTERNAL_DOWNLOADS]: The skill directs the user and the agent to depend on external software including the
patrolandpatrol_mcppackages from the official Flutter registry (pub.dev), and references documentation from the official Patrol site (patrol.leancode.co) and GitHub repository (leancodepl/patrol).
Audit Metadata