AI Task Spec Builder

Problem

When a user starts with a coarse request like “help me write a prompt” or “turn this into a task for an AI,” the hard part is often not execution but making the goal, inputs, constraints, output format, and acceptance criteria explicit. ai-task-spec-builder handles that front-end work: it converts vague requests into a task specification that another model can execute, instead of mistakenly treating final-deliverable tasks such as “optimize this” or “analyze this data” as prompt-generation tasks.

How it works

The skill narrows the request through a seven-step flow:
- Restate the raw request while preserving the user's intent and avoiding premature abstraction.
- Infer the real goal, for example interpreting “write a report” as producing an evidence-backed management summary.
- Classify the task type, including writing, research, coding, data analysis, product work, and decision support.
- Separate missing information into must provide and nice to provide, and make reasonable assumptions for low-risk gaps.
- Choose a response strategy, such as asking targeted questions, producing a best-effort prompt, generating multiple prompt variants, outputting a structured spec, or safely redirecting unsafe requests.
- Generate a copy-ready prompt with objective, context, inputs, assumptions, constraints, output format, quality criteria, and execution steps.
- Add an acceptance checklist so the result can be evaluated concretely rather than judged by vague adjectives.

Boundaries

Use it only when the user explicitly needs a prompt, template, or task specification. If the user asks for direct work such as “write my weekly report” or “fix this code,” perform the task directly instead of generating a prompt. For unsafe requests, it should not produce an actionable harmful prompt; it should redirect to risk assessment, compliance review, defensive guidance, or other safe framing. After delivery, it should not become the default mode for the rest of the conversation unless the user explicitly asks to refine, reuse, or create another prompt.