The Agent Skills Directory

[PROMPT_INJECTION]: The instructions designate certain task names (e.g., status, check, diff, list) as 'auto-approved' for agent execution based on naming conventions alone. This pattern attempts to influence the agent's internal permission and safety logic by asserting that these specific command names are non-mutating and safe by convention, which could potentially be leveraged to bypass manual review if a malicious task is given one of these conventional names.
[SAFE]: The skill provides legitimate best-practice guidance for avoiding interactive input in automation scripts, specifically citing security risks such as plain-text credential exposure due to pyinvoke stdin echoing and execution hangs in specific Python interpreter versions (3.14).
[SAFE]: External links and citations refer to well-known technology guidelines and official documentation from trusted organizations such as Microsoft (Azure CLI guidelines) and the Nix project.

invoke-task-conventions