AgentSkillsCN

operationalize-ambiguity

在尚未确立量化目标但利益相关方仍需具体、可审查的需求时,不妨将定性目标(如“值得信赖”“富有同理心”“公平公正”)转化为可落地执行的具体要求,而不必刻意为之,徒增虚假指标。

SKILL.md
--- frontmatter
name: operationalize-ambiguity
description: Turn qualitative goals (e.g., trustworthy, empathetic, fair) into actionable requirements without inventing fake metrics. Use in early discovery/product definition when stakeholders can’t yet commit to numeric targets but still need concrete, reviewable requirements.

Operationalize Ambiguity

Rules

  1. Allow qualitative terms only if you add one of:
    • Examples and non-examples (behavioral)
    • A qualitative rubric (what “good” vs “bad” looks like)
    • A study plan (usability test, red-team, stakeholder review)
    • An audit/logging plan
    • “TODO: metric plan” owned by a stakeholder + timeline
  2. Avoid fake precision:
    • Don’t invent “95% empathy” style metrics without a measurement plan and baseline.
  3. Treat “Evidence needed” as first-class output:
    • List the stakeholder decision, definition, or data needed to finalize the requirement.

Prompt snippet (copy/paste)

text
Qualitative terms are allowed, but each must include an Operationalization field:
- Examples / non-examples OR
- Rubric OR
- Study/audit/logging plan OR
- TODO metric plan (owner + decision needed).
Do not invent numeric targets unless a measurement plan and owner are specified.