Hi @DietrichGebert ,
Absolutely love the concept behind Ponytail. The "lazy senior dev" heuristic is a fantastic way to keep agent outputs clean and minimal.
I maintain a repository called vibe-coding-essentials , which focuses on workflow frameworks and anti-hallucination guardrails for AI-assisted development.
I've noticed that sometimes when an agent acts "lazy," it can take shortcuts by hallucinating variables or functions that don't exist in the codebase just to output less code. I'd love to submit a Pull Request to integrate a few of the core anti-hallucination guardrails from my framework directly into Ponytail's prompt heuristics.
It would effectively enforce a rule like: "You are a lazy senior dev, which means you write minimal code, but you NEVER invent functions that don't exist because debugging hallucinations takes too much effort."
Would you be open to a PR that adds these strict reality-check guardrails to the system prompts?
Hi @DietrichGebert ,
Absolutely love the concept behind Ponytail. The "lazy senior dev" heuristic is a fantastic way to keep agent outputs clean and minimal.
I maintain a repository called vibe-coding-essentials , which focuses on workflow frameworks and anti-hallucination guardrails for AI-assisted development.
I've noticed that sometimes when an agent acts "lazy," it can take shortcuts by hallucinating variables or functions that don't exist in the codebase just to output less code. I'd love to submit a Pull Request to integrate a few of the core anti-hallucination guardrails from my framework directly into Ponytail's prompt heuristics.
It would effectively enforce a rule like: "You are a lazy senior dev, which means you write minimal code, but you NEVER invent functions that don't exist because debugging hallucinations takes too much effort."
Would you be open to a PR that adds these strict reality-check guardrails to the system prompts?