ad-philosophy

To enforce think-before-code guardrails, apply universal agent rules that prompt explicit reasoning, surface ambiguities, and identify trade-offs before initiating complex edits. This ensures validated outcomes and reduces risk during non-trivial work.

Grounding before coding invokes a research pass to anchor decisions in real patterns and references. It improves documentation verification by ensuring definitions, decisions, and audits are captured accurately without leaking implementation details.

To apply minimalism and surgical changes, scope edits to the smallest necessary footprint with clear justification. This guardrail approach encourages well-scoped modifications and logs verification steps before task completion.

Yes, frontmatter presence with name and description is required for agent guardrails on non-trivial work. Optional scripts, references, and assets are also supported to ensure deliberate thinking and validated outcomes.

Use think-first guardrails for any non-trivial work requiring grounding, minimalism, and explicit documentation across tasks. Direct edits without guardrails risk unvalidated outcomes and unintended side effects in complex software engineering scenarios.