The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of external software from NPM (@mvanhorn/printing-press) and GitHub (github.com/mvanhorn/printing-press-library) using npx and go install commands. These sources are not in the verified safe list for this environment.
[REMOTE_CODE_EXECUTION]: The installation instructions for redfin-pp-cli and redfin-pp-mcp involve fetching and executing binary code from remote repositories at runtime.
[DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> argument that allows sending the results of any query (including property data and market trends) to an arbitrary external endpoint via HTTP POST.
[COMMAND_EXECUTION]: The skill facilitates the execution of the redfin-pp-cli binary with user-provided arguments. It also allows writing data to arbitrary file paths on the local system using the --deliver file:<path> parameter, which could be used to overwrite sensitive configurations.
[PROMPT_INJECTION]: The skill processes untrusted data from Redfin's internal endpoints. While categorized as a lower risk, the combination of processing external data with the ability to write files and make network requests increases the surface for indirect prompt injection attacks.

pp-redfin