Community研究與資料分析github.com

Jeremy-01/intake-first-planner

Codex skill for schema-first intake before planning and solver implementation

相容平台Claude CodeCodex CLI~Cursor
npx skills add Jeremy-01/intake-first-planner

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

說明文件

Intake-First Planner

Use this skill for projects that must not proceed from guessed parameters: physical layout planners, workstation or furniture arrangement tools, equipment placement, cable routing, schedule/resource optimization, simulation setup, and similar constraint-solving systems. Assume the user may be a complete beginner: explain what information is needed in plain language, provide fillable YAML, and avoid requiring them to understand implementation details.

Core Rule

Do not solve, optimize, recommend, or scaffold domain-specific logic until the required facts and scenarios are explicit enough to make assumptions unnecessary. If critical data is missing, output a "Missing Parameters" intake response and stop before implementation.

Workflow

  1. Classify the request:
    • Intake only: user wants requirements, schema, or prompt cleanup.
    • Build after intake: user wants code, tests, reports, or a reusable project.
    • Skill authoring: user wants a reusable/open-source skill or prompt package.
  2. Select the relevant reference:
    • General reusable workflow: read references/intake-workflow.md.
    • Spatial/layout planners: read references/spatial-layout-domain.md.
    • Three.js reports or 3D/visual output: read references/threejs-visualization.md.
    • Converting a one-off prompt into a reusable skill: read references/prompt-to-skill.md.
    • Skill packaging/open-source conversion: read references/skill-packaging.md.
  3. Produce or update a structured fact source, usually YAML or JSON.
  4. Gate execution:
    • If required facts are missing, return only the missing-parameter checklist, a suggested schema fragment, required scenarios, and explicit non-assumptions.
    • If facts are complete, implement the project using the fact source as authoritative input.
  5. For Python projects, prefer uv commands. Always include venv + pip recovery commands in handoff/recovery docs.
  6. Do not read old session JSONL, chat logs, or previous conversations unless the user explicitly asks for history recovery.

Output Patterns

For missing input, use this shape:

I need a few concrete measurements before I can solve this safely. You can answer by filling in the YAML below; unknown values may stay `null`.

**Missing Parameters**
- ...

**Suggested YAML**
```yaml
...
```

**Scenarios To Confirm**
- ...

**Not Assumed**
- ...

For implementation, create a project with:

  • pyproject.toml, src/, tests/, examples/, reports/, outputs/
  • A structured config example as the single fact source
  • Visual reports with top-view SVG/HTML and a Three.js 3D HTML viewer when geometry or layout is involved
  • Deterministic tests for parsing, transforms, constraints, scoring, and at least one scenario regression
  • HANDOFF.md, RECOVERY_README.md, and PROJECT_CONTEXT.md with both uv and venv + pip commands

Reusable Assets

  • schemas/intake_project.schema.yaml: generic schema template.
  • schemas/spatial_layout.schema.yaml: spatial/layout schema template.
  • references/threejs-visualization.md: default 3D visualization guidance.
  • examples/workstation_layout_request.yaml: example domain request.
  • examples/furniture_layout_request.yaml: room/furniture planning example.
  • examples/equipment_cable_routing_request.yaml: equipment placement and cable-routing example.
  • scripts/check_skill_package.py: lightweight package sanity checker.

相關技能