14 Production-Grade Agent Projects: Self-Healing Loops to Cost Kill-Switches
Full-Presence7590 · reddit · 2026-11-01
A follow-up Reddit post details 14 portfolio projects for agentic AI engineers, each with build steps and rationale: an async orchestrator handling 100+ concurrent tasks, a context budget assembler, strict Pydantic tool routing with auto-retry on validation failure, a 3-tier memory engine with TTL eviction, a self-healing ReAct loop that parses its own stack traces, typed LangGraph multi-agent swarms, an async Slack human-approval gate, a trajectory-eval CI gate that replays 100 trajectories, OpenTelemetry tracing, an inter-agent sanitization proxy against cross-agent prompt injection, a sandboxed E2B/gVisor code executor, a $0.50 cost kill-switch middleware, an API-to-headless-browser fallback, and a 5% shadow traffic comparator for diffing agent topologies.
More from coding & agent
- Screenshot shows AI agent autonomously requesting an account via headless signup — repligate · 2026-11-01
- Hermes browser extension v0.4.0 ships 27 browser tools with opt-in control — Teknium · 2026-11-01
- Shreya Shankar's group publishes 5 rules for readable bulk AI text — sh_reya · 2026-11-01
- ChatGPT and Codex to share one token-based usage allowance from December — TinfoilTricorn · 2026-11-01
- Teachable Figures: open-source agent skill for interactive 2D/3D lesson figures — yshan2u · 2026-11-01
- A week of AI coding made me feel like an 'expediter', not a programmer — sloppenheimer · 2026-11-01