Sign InOpen Brain
AI EngineerVideoSource Linked

Agents Without Code: Skills, YAML, and Filesystems Replaced Python — Philipp Schmid, Google DeepMind

A GitHub review agent is rebuilt with progressively less orchestration code, ending as instructions and setup files around general tools. Evals still remain the builder’s job.

AI Engineer · Sep 14, 2026
Open Source Open MarkdownOpen JSON
Source Summary

The talk rebuilds one pull-request review agent three ways: a Python tool loop, an agent framework and a hosted filesystem agent. In the third version, the source directory disappears; the remaining agent definition is an **AGENTS.md** file plus a small GitHub CLI installer.

Practical Implication

For new coding agents, start with general capabilities such as Bash, files and established CLIs before writing narrow tool wrappers. Move domain rules into skills and files, let the runtime manage loops and state, and keep **outcome evals** under your control.

Agent-Ready Context
The talk rebuilds one pull-request review agent three ways: a Python tool loop, an agent framework and a hosted filesystem agent. In the third version, the source directory disappears; the remaining agent definition is an **AGENTS.md** file plus a small GitHub CLI installer.

For new coding agents, start with general capabilities such as Bash, files and established CLIs before writing narrow tool wrappers. Move domain rules into skills and files, let the runtime manage loops and state, and keep **outcome evals** under your control.

The simplified version depends on a hosted sandbox, server-managed state, credential injection and capable models. Removing orchestration code does not remove the need for network restrictions, clean tools, verification or explicit behavioral rules.
Connected Context · Feed7 Judgment

This consolidates the filesystem-and-skill pattern into a design rule: begin with general tools and remove custom orchestration only when the runtime can safely supply state, isolation and credentials. It confirms the Hugging Face and Vercel examples, while narrowing their apparent simplicity by keeping evaluation, verification and security controls outside the reduced agent definition.

How We Solved Agent Building — Andrew Qu, VercelVercel reports the same shift from prescriptive pipelines to a sandboxed filesystem agent, providing workload-specific eval evidence for the architecture advocated here.How I automate my own job at Hugging Face using agents — Niels Rogge, Hugging FaceThe Hugging Face workflow is a bounded production example of replacing custom workflow code with a mature CLI, one skill and a sandbox.From coding to Knowledge work agents — Karan Vaidya, ComposioComposio supplies an important implementation boundary: permissions, action records and irreversible-action checks must remain enforced outside the instructions even when orchestration code disappears.500 Skills, Zero Fine-Tuning: LinkedIn's Playbook for AI Agents — Ajay Prakash, LinkedInLinkedIn shows how the proposed skill-first approach changes at catalog scale: capabilities must be progressively discovered and governed rather than all loaded into the agent at once.
Context Map
agentcoding#harness-engineering#skills#tool-use
Uncertainty
The simplified version depends on a hosted sandbox, server-managed state, credential injection and capable models. Removing orchestration code does not remove the need for network restrictions, clean tools, verification or explicit behavioral rules.