Agent Experience

Designing software, repos, docs, SDKs, and feedback loops for AI agents as a first-class consumer audience — the agent-facing analog of UX and DX. Grounding lives in skill.json; this file is runtime routing only.

Produces: an AX review/design/debug report for the chosen surface; or, when the job is to build / harden / measure, a hand-off to the right implementation arm.

Core principle

Treat the agent as an audience with its own ergonomics: machine readability, context budget (W6), deterministic action, stable contracts, and evidence-driven scaffolding (never /init; W9). The good-shaped path is the one a stateless, context-window-bound agent can find, parse, and act on.

Two things this skill does