The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute local CLI tools within the development environment, specifically dart analyze for identifying code issues and dart format for applying style corrections. These operations are intended for developer-led workflows and are restricted to the local filesystem scope.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository data, creating a potential vector for indirect prompt injection.\n
Ingestion points: The agent reads and parses Dart source code files (.dart) and project metadata (pubspec.yaml), which could contain malicious content designed to influence the AI's behavior.\n
Boundary markers: The instructions do not mandate the use of specific delimiters or isolation markers when the agent reads and reflects file content into its context.\n
Capability inventory: The agent possesses the capability to modify project files (via dart format) and execute shell commands derived from project analysis results.\n
Sanitization: The skill relies on the standard output of the Dart SDK and does not include explicit logic to sanitize or filter strings extracted from repository files before they are processed by the LLM.\n- [EXTERNAL_DOWNLOADS]: The skill includes procedures to verify the presence and version of the Dart/Flutter toolchain. However, it explicitly prohibits the automatic or silent installation of external software, dependencies, or upgrades, ensuring that any missing prerequisites are handled manually by the user.

codeguard-dart