The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install CLI and MCP components from external sources linked to the vendor mvanhorn. This includes the use of go install for repositories on GitHub and npx for packages on NPM.
[DATA_EXFILTRATION]: A built-in feature of the redfin-pp-cli allows output to be delivered to a webhook:. This enables the agent to send processed data to any external endpoint, creating a potential vector for exfiltrating sensitive information or search results to untrusted locations.
[INDIRECT_PROMPT_INJECTION]: The skill is intended to process external real estate data which may contain untrusted content.
Ingestion points: Data is ingested from external Redfin API endpoints via commands like homes, listing, and market as described in SKILL.md.
Boundary markers: The instructions do not define boundary markers or delimiters to isolate untrusted listing data from the agent's instructions.
Capability inventory: The skill environment allows file writes and network POST requests through the CLI's --deliver flag, and uses the Read Bash tool for execution.
Sanitization: There is no documentation of sanitization or validation of the ingested content to prevent embedded instructions from influencing agent behavior.

pp-redfin