The Agent Skills Directory

[SAFE]: The skill implements a strict, phase-gated workflow for C# development, which enhances security by requiring explicit user confirmation before proceeding with code changes or commits. It emphasizes that urgency or triviality does not waive these safety gates.
[DATA_EXFILTRATION]: The skill includes defensive instructions to refuse staging or committing sensitive files such as .env, credentials.json, and *.pem files, and it mandates that all commits must be explicitly directed by the user with named staging only.
[COMMAND_EXECUTION]: While the skill uses shell commands for building, testing, and smoke-checking the application, these are standard development tasks performed on the local project code under user oversight. The 'App Smoke-Check' specifically instructs the agent to check startup logs and health endpoints, which is a legitimate verification step.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external requirements but applies a multi-phase review and clarification process (Phase 1 and 2) which serves as a functional boundary against unintended instruction following by forcing the agent to extract and verify criteria before implementation.

dotnet-dev