{"author":"sudhendra1","children":[{"author":"sudhendra1","children":[],"created_at":"2026-08-28T19:45:18.000Z","created_at_i":1787946318,"id":49483331,"options":[],"parent_id":49483173,"points":null,"story_id":49483173,"text":"I built this after getting paged one time too many for an agent that leaked a Slack token to Claude. It's an OpenAI-compatible proxy plus MCP-tool hooks that block credentials, prompt injection, and out-of-policy tool calls before they hit the provider. Every call is hash-chained so you can prove what the agent actually did. Curious what breaks first for people \u2014 happy to debug live.","title":null,"type":"comment","url":null},{"author":"sigy","children":[],"created_at":"2026-08-28T20:51:39.000Z","created_at_i":1787950299,"id":49484107,"options":[],"parent_id":49483173,"points":null,"story_id":49483173,"text":"If you want something much lighter, and focused around harness tool calls, check this out: https://crates.io/crates/coding-tools","title":null,"type":"comment","url":null},{"author":"LuD1161","children":[],"created_at":"2026-08-28T21:30:26.000Z","created_at_i":1787952626,"id":49484458,"options":[],"parent_id":49483173,"points":null,"story_id":49483173,"text":"I've been working on something similar , policy guardrails for coding agents, with OPA backed policies and os-native sandbox - https://github.com/LuD1161/agentjail","title":null,"type":"comment","url":null},{"author":"rob_b","children":[],"created_at":"2026-08-28T21:32:05.000Z","created_at_i":1787952725,"id":49484482,"options":[],"parent_id":49483173,"points":null,"story_id":49483173,"text":"Another example of why vibe coding is the worst advancement ever.","title":null,"type":"comment","url":null}],"created_at":"2026-08-28T19:29:20.000Z","created_at_i":1787945360,"id":49483173,"options":[],"parent_id":null,"points":22,"story_id":49483173,"text":null,"title":"Show HN: Conduct, open-source guardrails for LLM and MCP tool calls","type":"story","url":"https://github.com/sseshachala/conductai"}