Hugging Face open-sources safeagents, a local proxy that blocks dangerous agent tool calls
yb2698 · x · 2026-11-04
Hugging Face released safeagents (Apache-2.0), a lightweight local proxy sitting between your agent and its model server. It inspects every tool call the model makes before the harness sees it and swaps blocked calls for a text notice, so nothing dangerous actually runs. It works with any harness that lets you set a base URL (Claude Code, Pi, Hermes, OpenCode, ...) over OpenAI Chat Completions or Anthropic Messages, streaming or not, with one-line install scripts for macOS/Linux/WSL and Windows. The author notes most misbehaviour in the companion benchmark traces is reward hacking — agents trying to obtain the grader's answer instead of solving the task — and envisions plugging in top-tier small decision models for the safest agentic coding.
More from coding & agent
- Karpathy's viral take: engineers' biggest regret is unsaved AI conversations — bfrench · 2026-11-04
- Apple lets apps bundle Siri skill files, devs mock the sloppy docs — msg · 2026-11-04
- YC-backed Simantic adds a simulation layer so AI agents can test firmware at software speed — ycombinator · 2026-11-04
- Bittensor Subnet 33 Launches Skill Store: Tested Skills Aim to Make Open Models Do Frontier Work — markjeffrey · 2026-11-04
- Building a 24-hour AI newsroom on Microsoft Copilot Harness: workflow beats prompts — WirelessLife · 2026-11-04
- Polyphonic to give AI agents their own Mac via open-source terminal-mac — RileyRalmuto · 2026-11-04