The Agent Skills Directory

[COMMAND_EXECUTION]: The skill recommends that users modify their .claude/settings.json file to include wildcard execution permissions for the dotnet CLI. Specifically, it suggests auto-allowing Bash(dotnet run *) and Bash(dotnet tool *). This configuration bypasses the user confirmation step for commands that can execute arbitrary project code or install third-party software, creating a risk of autonomous malicious activity if the agent is provided with compromised project files or malicious instructions.\n- [PROMPT_INJECTION]: The skill promotes workflows that involve processing untrusted external data, such as CI failure logs and GitHub PR review comments, to automate debugging and refactoring. This ingestion of external data without proper safeguards creates a risk of indirect prompt injection.\n
Ingestion points: The skill instructs the user to send CI logs to a 'build-error-resolver' and use PR comments to drive code review automation.\n
Boundary markers: There are no instructions provided to wrap these external inputs in delimiters or to include 'ignore' commands to prevent the agent from obeying instructions embedded in the logs or comments.\n
Capability inventory: The skill explicitly encourages granting the agent broad shell execution capabilities for the dotnet ecosystem with pre-approved permissions.\n
Sanitization: The instructions do not recommend any sanitization, filtering, or validation of the content received from external logs or reviews.

workflow-mastery