The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md and references/quickdup.md recommend installing the QuickDup CLI by piping remote scripts directly from a third-party GitHub repository into shell interpreters, a high-risk execution pattern.
Evidence: curl -sSL https://raw.githubusercontent.com/asynkron/Asynkron.QuickDup/main/install.sh | bash in SKILL.md.
Evidence: iwr -useb https://raw.githubusercontent.com/asynkron/Asynkron.QuickDup/main/install.ps1 | iex in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill downloads installation scripts and Go packages from an external repository (github.com/asynkron/Asynkron.QuickDup) that is not recognized as a trusted vendor or well-known service.
[COMMAND_EXECUTION]: The skill performs several shell operations, including tool installation (go install, curl, iwr) and tool execution (quickdup) with parameters determined at runtime based on the source code structure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted source files from the repository without sanitization, which could contain instructions meant to manipulate the agent.
Ingestion points: C# (.cs) and Razor (.razor) files, and the duplication report output from the QuickDup tool.
Boundary markers: None; the skill lacks delimiters or instructions to distinguish code content from potential embedded agent commands.
Capability inventory: The agent has shell access to execute CLI tools and modify repository files during refactoring workflows.
Sanitization: No sanitization or verification of the file content is performed before processing.

quickdup