Back to directory

Hermes Agent Idea Workflow

A set of Hermes Agent skills that turn a rough product idea into a design document, an implementation specification and a single build-handoff file another agent can implement from, before any coding starts.

Hermes Agent Idea Workflow

Idea Workflow covers the product-thinking phase that comes before a build, separating it from implementation thinking so an agent can capture what a user actually wants, clarify the intended experience, produce a design document and an optional UI design brief, write an implementation specification, and package everything into one handoff file. It explicitly does not execute or build anything itself; it produces the source material meant to feed a separate build workflow, recommending a companion project as the next step for turning that handoff into a coding plan.

A lite mode captures a quick idea as a single note plus an index entry, meant for sketches or small utilities that only need enough structure to revisit later. A full mode is for a more serious idea and produces a numbered sequence of files: idea capture, a design document, an optional UI design brief, an implementation specification, the agent build handoff, and a spec review gate, with the UI stage skippable while keeping the rest of the sequence intact.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.