AGENT SKILL

calibrate-audience

Establish audience skill level and app shape early, then right-size the architecture as a conscious choice.

License 许可
MIT
最近更新
2026-11-07
Stars
★ 1
Forks
0

01 前言 Front Matter

name
calibrate-audience
description
Establish audience skill level and app shape early, then right-size the architecture as a conscious choice.

03 SKILL.md 正文

Calibrate Audience

Purpose

Establish who the work is for and what shape it is, early and in plain language, so the architecture is right-sized as a deliberate, recorded choice.

Procedure

  1. Infer from repository evidence first: existing runtime, entrypoints, UI, services and docs. Ask only the smallest set of questions evidence cannot answer.
  2. Establish skill level on a simple scale: normal_folk (little or no coding background) to developer. This sets explanation depth and jargon.
  3. Establish app shape in normal-folk terms:
    • browser-only (runs entirely in the visitor's browser, nothing to run on a server);
    • full-stack (a browser part plus a server part that stores or processes data);
    • service, CLI, library, mobile, desktop or game as evidence indicates.
  4. Recommend language by role, not fashion: match the runtime already present, the team's stated skill, and the app shape.
  5. Produce an explicit right-sizing recommendation: the smallest sufficient architecture, what is deliberately excluded and why, and the conditions that would justify revisiting it.

Output

Record to PROJECT_PROFILE.toon:

audience:
  skill_level: normal_folk | developer
  app_shape: browser_only | full_stack | service | cli | library | mobile | desktop | game
  language_by_role: []
  right_sizing:
    smallest_sufficient: one line
    deliberately_excluded: []
    revisit_when: []
    bought_in: true

Rules

  • Right-sizing is never silent. State exclusions, get buy-in, and record the decision (ADR when durable) before proceeding.
  • Explain full-stack versus browser-only for a non-technical audience.
  • Prefer evidence over questions; prefer the smallest useful question set.
  • Re-run when audience or app shape changes materially.

04 同仓库更多 Skills