The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data by reading GitHub issues via gh issue view and pull request comments via gh pr view and the GitHub API. This creates a surface for indirect prompt injection where an attacker could embed malicious instructions in an issue or comment to influence the agent's actions during the reproduction or review phases.
Ingestion points: gh issue view, gh pr view, and gh api calls to the comments endpoint.
Boundary markers: Absent. The instructions do not define delimiters or provide specific warnings to the agent to disregard instructions found within the issue or comment data.
Capability inventory: The agent is authorized to execute shell commands (dart, gh), modify the local filesystem, and create or update pull requests.
Sanitization: Absent. There is no mention of sanitizing or validating the input from GitHub before processing it.
[COMMAND_EXECUTION]: The skill relies on the execution of several CLI tools to perform its tasks.
Evidence: Commands such as dart pub global activate, dart run, dart test, dart analyze, and various gh (GitHub CLI) commands are used to manage dependencies, run tests, and interact with the repository.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to download and install the project's CLI directly from its Git repository.
Evidence: Phase 2 contains the command dart pub global activate --source git https://github.com/leancodepl/patrol.git. This is a standard development practice within the vendor's own repository environment.

fix-issue