Building your own agent harness: why banks and hospitals skip off-the-shelf coding agents
TejasKumar_ · x · 2026-11-18
An Insecure Agents interview with Tejas Kumar (author of Fluent React) on his AIE CODE workshop and building the Daiku personal-agent harness:
- Banks, insurers, and hospitals build their own harnesses rather than use off-the-shelf coding agents
- Live-demo risk: a "Delete everything" prompt actually deleted everything on stage
- Guardrails after the rm -rf era: allow lists, ask-first tools, Face ID before sensitive actions
- Key insight: "can the agent call this tool" vs "should it call it right now" are different questions; agent actions need tracking and logging
Useful for designing production agent permission boundaries.
More from coding & agent
- Academic Research Skills for Claude Code hits 57.5k GitHub stars — adnan_hashmi · 2026-11-18
- Google paper: gated self-improving agent harnesses cut token cost 10x and transfer across models — omarsar0 · 2026-11-18
- Replan projects via MCP: a two-step prompt pattern for Codex and Claude Code — sfoooooooooooooooooo · 2026-11-18
- CallForMe plugin lets ChatGPT phone US businesses for quotes and bookings — jakemarsh · 2026-11-18
- Probe Volumes lands in PlayCanvas engine with GPU re-baking and day/night setups — willeastcott · 2026-11-18
- Meme: developer too busy watching his coding agent migrate databases to care about GTA6 — vmg · 2026-11-18