herdr

When intent or consequential choices are unclear — resolve goals, boundaries, contracts, and an approved direction before building.

When maintaining, assessing or redesigning the agent harness itself — skills, agents, extensions, hooks, tools, setup, docs and tests — or absorbing upstream…

Creating, revising, or cleaning a skill — keep one useful mindset, a discriminating trigger, and only guidance that earns its context cost.

When creating, restyling, or holistically reviewing web UI — set a coherent aesthetic direction, coordinate its design disciplines, and build or assess the…

When deciding whether a frontend interaction benefits from motion, or designing or reviewing state, enter/exit, and gesture motion — make movement purposeful…

When refining component surfaces, elevation, optical alignment, icons, or finish after the structure is sound — remove distracting detail while preserving the…

When writing, refactoring, testing, or configuring Rust — follow Bryan's toolchain, error-handling, crate, API-design, and structure preferences and the…

When defining, changing, or reviewing UI color tokens, themes, contrast, gamut, or light/dark/increased-contrast appearances — preserve semantic meaning and…

When the user explicitly wants a guided lesson or course — teach a focused session in chat by default, with a persistent learning workspace only on request.

When structure is the problem — improve an existing architecture without changing its goals, or rethink it from first principles when the user explicitly asks.

When asked for a commit, its message, or a push — preserve staged intent and verified scope; execute only explicitly authorized operations.

When shaping or reviewing a frontend's spatial structure—grouping and aligning content, setting reading order, adapting a component to its container, or…