The Agent Skills Directory

[SAFE]: The skill's primary function is intent classification and routing. It does not perform direct network requests, credential access, or shell command execution. It explicitly forbids autonomous Git operations (commits/pushes), ensuring user oversight is maintained for code changes.
[PROMPT_INJECTION]: The skill ingests user prompts to classify engineering tasks. While this is a potential surface for indirect prompt injection, the skill acts as a pass-through orchestrator and implements defensive gates, such as explicit 'Risk signals' that force higher-scrutiny modes and user confirmation checkpoints.
[COMMAND_EXECUTION]: All functionality is delegated through official tool interfaces (Skill and AskUserQuestion). The skill does not use subprocesses, system shells, or dynamic code execution to perform its logic.

ship