Add Human Approval Audit to Agents
Individual_Chapter62 · reddit · 2026-07-15
The author built a tool called Approv that pauses agent execution on high-risk actions (e.g., refunds, account changes, writing to production DB) and requests human approval/denial via WhatsApp.
Each approval is recorded as an immutable, signed entry. The author seeks feedback on how others handle human-in-the-loop approval for agents and whether such audit trails are truly necessary.
More from coding & agent
- OpenWiki adds Gemini AI Studio and Vertex AI support for codebase docs — BraceSproul · 2026-07-22
- OpenWiki adds Gemini AI Studio and Vertex AI support plus Gemini 3.6 Flash — BraceSproul · 2026-07-22
- Kimi K3 feels slower than K2.7, but stronger on long coding jobs and refactoring — Far-Presence2711 · 2026-07-22
- Poolside launches Laguna S 2.1 with 118B parameters and 8B active per token — Madisonkanna · 2026-07-22
- OpenWiki adds Gemini AI Studio, Vertex AI, and new Flash models — BraceSproul · 2026-07-22
- A Forward Deployed Engineer job really has three stages: audit, evals, deploy — blaizedsouza · 2026-07-22