The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by gathering stack traces, error messages, and git diff outputs from the user's environment. This data represents a potential attack surface where malicious instructions could be embedded to influence the agent's behavior.
Ingestion points: The skill instructs the agent to gather "Full error message and stack trace" and "Recent changes: git diff" in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters defined to isolate or sanitize the ingested text from the agent's core instructions.
Capability inventory: The skill has the capability to execute shell commands and build tools including flutter, git, gradlew, and pod.
Sanitization: No sanitization process is described for the logs or code snippets before they are processed by the agent.
[COMMAND_EXECUTION]: The skill facilitates the execution of local development and diagnostic commands through the agent's shell capabilities.
Evidence: Instructions include running flutter --version, flutter doctor -v, flutter run --verbose, git diff, and cleaning build artifacts via gradlew and pod (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill uses standard package management commands to resolve build issues, which involves downloading dependencies from well-known official registries.
Evidence: The skill directs the agent to use flutter pub get for Dart packages and pod install for iOS dependencies (SKILL.md). These operations fetch code from established registries (pub.dev and cocoapods.org).

flutter-debugging