Split code writing and review across sub-agents to dodge safety blocks with Fable

When an AI writes code, its built-in safety filters can suddenly stop it mid-task if the code looks suspicious. This workflow fixes that by giving the writing job to one agent and the review job to a separate Fable-powered sub-agent. Splitting the roles keeps the pipeline running without unwanted interruptions.

AI models like Claude have internal safety filters that can flag certain code patterns — system commands, file operations, or anything that looks potentially harmful — and refuse to continue. This is frustrating when you're building an automated pipeline that needs an AI to write, test, and fix code in a loop. The Reddit post describes a practical fix: separate the 'writer' agent from the 'reviewer' agent entirely.

The reviewer sub-agent, built on the Fable model, only sees finished code and evaluates it — it's not generating anything sensitive itself, so the safety filter rarely triggers in that context. This role-separation trick makes multi-agent coding pipelines more stable and lets complex automation tasks complete without manual intervention. Fable is one of Anthropic's latest models, known for strong reasoning over code, which makes it a good fit for the reviewer role.

Key points

  • Safety filters in AI models can block code generation mid-task; splitting roles avoids this
  • A main agent writes the code; a Fable sub-agent reviews it separately
  • The reviewer agent operates in a 'read and evaluate' context, which rarely triggers safety blocks
  • This pattern improves stability in automated code-generation and self-fix loops
  • Fable is Anthropic's latest model with strong code reasoning, well-suited for reviewing

Quick term guide

safety filters
Built-in rules that stop an AI from producing harmful or dangerous outputs.
safety filter
An automatic rule inside an AI model that stops it from producing content it judges as harmful or risky.
sub-agent
A smaller AI agent that handles one specific part of a larger task, running separately from the main agent.
AI models
The core brain or underlying program that powers an artificial intelligence tool.
AI model
A program that can understand prompts and produce text, code, or answers.
multi-agent
A setup where several AI agents each handle a different subtask and work together to complete a larger goal.
automation
A way to make repeated work happen without doing every step by hand.
reasoning
The ability of the AI to think through complex steps to find a solution.
Read original