Open-Source Agent Skill 'Foreman' Helps Coding Agents Design & Review Tools Safely
blair_hudson · reddit · 2026-07-05
The author open-sourced Foreman, an agent skill that helps coding agents like Claude Code and Codex identify often-overlooked security issues when designing or reviewing tools (including MCP and native framework tool calls). It checks for capability boundaries, authorization, side effects, idempotency, confused deputy risks, and failure modes. The author previously authored a book on generative AI tool security and topped Amazon Australia's bestseller list.
Foreman is currently operational. The author states the next step is to figure out how to evaluate this skill (exploring eval methods for agent skills) and is soliciting feedback and ideas from the community in their post.
Related event: Open-Sourcing Foreman to Help Coding Agents Safely Review Tools(2 posts)→
More from coding & agent
- Hugging Face wraps 6-month Training Agents series with open recordings and code — SergioPaniego · 2026-10-26
- Agent verified 'success' after double-voting: a lesson in harness vs model bugs — TejasKumar_ · 2026-10-26
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26