The Agent Skills Directory

[DATA_EXFILTRATION]: The CLI includes a --deliver webhook:<url> flag that allows the agent to POST the output of any command to an arbitrary external endpoint. This provides a direct mechanism for exfiltrating data, such as search results, price logs, or local configuration, to a remote server controlled by an attacker.
[DATA_EXFILTRATION]: The feedback command allows sending data to a remote URL defined by the RENO_GOAT_FEEDBACK_ENDPOINT environment variable, which serves as an additional egress point for local data.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the installation of binaries via go install from a remote GitHub repository (github.com/mvanhorn/printing-press-library) and via npx from the NPM registry (@mvanhorn/printing-press-library). This results in the execution of external code that is not verified by the platform's standard security controls.
[PROMPT_INJECTION]: The skill is designed to ingest and process data from 33+ active external retail sources (e.g., Ferguson, West Elm, Rejuvenation). This creates an indirect prompt injection surface where malicious instructions embedded in product descriptions or metadata on these sites could potentially influence the agent's behavior during data enrichment or comparison tasks.
[COMMAND_EXECUTION]: The skill requires the execution of a local binary (reno-goat-pp-cli) and provides instructions for its installation and use via shell commands. The ability to write command output to files via the --deliver file:<path> sink adds a risk of unauthorized file writes in the local environment.

pp-reno-goat