agent-operating-pattern

Standardize agent workflows by applying a six-step context-driven flywheel: load context, query data, reference examples, execute, save to dashboard, and feed back. This enforces data loading rules and quality gates to prevent hallucinations and ensure traceability across all tasks.

Structure an agent execution playbook using a standardized six-step cycle applied across domains like marketing, sales, content, engineering, and ops. Document the execution cycle clearly at the skill root, utilizing optional scripts, references, and assets directories to support consistent task execution.

The six-step agent flywheel handles self-improvement through a feed back loop that follows task execution and saving outputs to a dashboard. This iterative learning mechanism captures results from completed tasks to continually enhance and refine future agent performance.

Yes, you can apply a context-driven workflow pattern to both engineering and marketing agent tasks. The standardized six-step cycle is designed to operate across domains, ensuring consistent data-backed outputs, accurate context loading, and continual learning regardless of the specific operational area.

Yes, implementing the agent operating pattern requires frontmatter with a name and description at the skill root. You also need a clearly documented execution cycle, alongside optional scripts, references, and assets directories to fully support the workflow execution and data querying steps.

Agent task workflows lack consistent context and data-backed outputs when they fail to enforce standardized data loading rules and quality gates. Implementing a six-step flywheel ensures every task loads documents, queries data, and references examples before execution to prevent traceability gaps.