SuperAPI AI Directory

purpose-discovery

Agents & models

View repository

Updated 10-16 01:33

About this skill

Purpose Discovery for Skills Development

Clarify the user's true objective and generate an optimized specification before skill generation.

Use This Skill When

  • A user requests Agent Skills development but the request is vague.
  • Key information (domain, audience, workflows, integrations) is missing.
  • Requirements need structuring before generation begins.

Workflow

Phase 1: Information Sufficiency Check

Evaluate the user's input against these 8 required elements:

# Element Question to ask if missing
1 PURPOSE What decision, workflow, or outcome should this skill enable?
2 DOMAIN What professional domain or subject area?
3 AUDIENCE Who will use this skill?
4 SCOPE Single skill or suite? How many sub-skills?
5 WORKFLOWS What are the main task types or phases?
6 INTEGRATIONS Does it need MCP tools, databases, or APIs?
7 REFERENCE MODEL Is there an existing skill group to reference?
8 QUALITY CRITERIA What defines success?

Sufficiency rule: 5+ elements clear → proceed to Phase 3. <5 elements clear → enter Phase 2 dialogue.

Phase 2: One-Question-at-a-Time Dialogue

Rules:

  • Ask exactly one question per turn. Never batch multiple questions.
  • Prefer closed or multiple-choice questions over open-ended.
  • After each answer, re-check sufficiency (5+ elements met?).
  • Maximum 8 rounds. After 8, proceed with stated assumptions.

Priority order (ask most impactful gaps first):

  1. PURPOSE → 2. DOMAIN → 3. WORKFLOWS → 4. SCOPE
  2. AUDIENCE → 6. INTEGRATIONS → 7. REFERENCE MODEL → 8. QUALITY CRITERIA

Phase 3: Structured Prompt Generation

Compile into optimized specification:

# Agent Skills Development Specification

## Objective
[One-sentence purpose statement]

## Domain & Audience
- Domain: [domain]
- Primary users: [audience]

## Architecture
- Type: [Single / Suite]
- Sub-skills: [list]
- Custom Agents: [list with roles]

## Workflow Phases
| Phase | Sub-skill | Description | Gate |
|-------|-----------|-------------|------|
| 0 | [name] | [desc] | ⏸️/auto |

## Integrations
- MCP: [list or none]
- Databases: [list or none]

## Reference Model
- Based on: [existing group or none]

## Quality Criteria
- [criteria]

## Assumptions
- [any assumptions from incomplete info]

Phase 4: User Approval ⏸️

Present specification and wait for approval.

  • Approved → pass to skill-scaffolder.
  • Modifications → update and re-present.

Deliverables

  • Structured specification (input for skill-scaffolder).
  • results/skill-spec.md (saved for reference).

Quality Gates

  • PURPOSE is clear and actionable.
  • At least 5/8 elements explicitly addressed.
  • Specification uses standard template format.
  • User approved specification before generation begins.
  • Assumptions are explicitly stated.

If any gate fails: return to Phase 2 dialogue.

Gotchas

  • Do not ask multiple questions at once. Strictly adhere to one question, one answer
  • The user's first input is a "request," not a "specification." The specification is structured through dialogue
  • Do not determine SCOPE while PURPOSE remains ambiguous. If the purpose is unclear, you will misjudge the number of skills
  • If information is not gathered within 8 rounds, proceed by clearly stating assumptions. Do not fall into an infinite loop
  • Even when referencing an existing suite, customize it to match the user's purpose

Validation Loop

  1. Execute Sufficiency Check
  2. Check: Are 5/8 elements clear, and is PURPOSE specific?
  3. Insufficient → Additional hearing via one question, one answer
  4. Proceed to skill generation only after user approval