aura-guard

4 ms·

1. 1.▲
   by aura-guard 8mo ago

   API-level spend caps solve the "how much" problem but not the "why." The agent still loops 50 times before hitting the limit. You just lose $50 instead of $200. The missing layer is detection inside the agent loop itself
2. 2.▲
   by aura-guard 8mo ago

   This is exactly the problem. The blast radius question is real. One pattern I've been exploring is a deterministic governor that sits between the agent loop and the tools. Before any tool executes, it checks: is this a repeat? Is this
3. 3.▲
   by aura-guard 8mo ago

   One thing I haven't seen discussed much is enforcement at the tool-call boundary, specifically the moment between 'the model wants to call this tool' and 'the tool actually executes.' Most safety approaches focus on
4. 4.▲
   by aura-guard 8mo ago

   Update: Now published to PyPI (v0.3.2). Install: pip install aura-guard Canonical repo: https://github.com/auraguardhq/aura-guard
5. 5.▲
   by aura-guard 8mo ago

   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 inst
6. 6.▲
   Show HN: Aura Guard – stop agent tool loops and duplicate side effects(github.com)

   3 points by aura-guard 8mo ago|2 comments