While the standard webpage retrieval tools in Claude Code/Codex do some sanitization and filtering, out of the box Pi defaults to curl. This is elegant but wasteful.
I threw together scurl as a simple wrapper over curl to provide agent friendly markdown from web pages, and to limit basic secret exfiltration patterns. I intend to add prompt injection detection/mitigation down the line.
While the standard webpage retrieval tools in Claude Code/Codex do some sanitization and filtering, out of the box Pi defaults to curl. This is elegant but wasteful.
I threw together scurl as a simple wrapper over curl to provide agent friendly markdown from web pages, and to limit basic secret exfiltration patterns. I intend to add prompt injection detection/mitigation down the line.