Show HN: Aura Guard – stop agent tool loops and duplicate side effects
github.comI built Aura Guard after repeatedly seeing tool-using agents burn money the same ways: looping search calls, retrying 429/timeouts forever, and double-firing side effects (refund twice / duplicate emails).
Aura Guard is a small Python middleware that sits between an agent loop and its tools. Before a tool runs, it returns a deterministic decision (ALLOW / CACHE / BLOCK / REWRITE / ESCALATE / FINALIZE). No LLM calls inside the guard, just counters + signatures + similarity checks.
Quick demo (no API key):
pip install git+https://github.com/auraguarddev-debug/aura-guard.git aura-guard demo
Feedback welcome, especially on “rephrase & retry” detection thresholds (currently overlap coefficient 0.60, repeat threshold 3). Author here. Happy to answer questions.
If you’ve dealt with “rephrase & retry” tool loops or duplicate side effects (refund/email twice), I’d love feedback on the heuristics and thresholds.
Quick demo is aura-guard demo after install. Repo has a benchmark harness + JSON reports. Update: Now published to PyPI (v0.3.2).
Install: pip install aura-guard
Canonical repo: https://github.com/auraguardhq/aura-guard