MyProject Manager

Core Rules (Do Not Skip)

• •Perform environment survey first, including OS, CPU, GPU, RAM, disk, and existing toolchain.
• •Perform online research for relevant open-source projects/models and cite sources with dates.
• •Build a full project plan before implementation: phases -> steps -> verification per step.
• •Require verification pass before moving to the next step.
• •Enforce offline-first: models must be local; no runtime downloads.
• •Enforce portability: project can be packaged and shared; receiver can run locally.
• •Enforce separation and decoupling for long-term maintainability.
• •Require GPU acceleration for model runtime when supported; CPU-only is considered a failure unless the model cannot run on GPU.

Workflow (Strict Order)

1. •Environment Survey

• •Collect current machine specs and toolchain.
• •Output minimum / recommended / high-performance requirements.
• •Decide run mode based on hardware constraints.
• •If needed, use scripts/check_env.py to summarize specs.

1. •Online Research (Mandatory)

• •Search and list candidate open-source projects/models.
• •Record sources with access date.
• •Identify compatibility risks between new and old models.
• •Use references/source-log-template.md.

1. •Master Plan (Before Any Build)

• •Produce a phased plan with small steps.
• •Attach verification criteria to each step.
• •Use references/plan-template.md.
• •Use references/phase-atomic-template.md for phase -> subphase -> atomic steps.

1. •Architecture & Dependency Isolation

• •Define module boundaries and interfaces.
• •Choose isolation strategy (venv / container / per-model env).
• •Use references/conflict-matrix-template.md to record conflicts.

1. •Model Local Management

• •Enforce local model storage layout.
• •Provide auto-download + manual-download paths.
• •Provide update prompt flow; do not auto-update by default.
• •Require GPU acceleration if supported and validated; CPU-only is allowed only when the model cannot run on GPU.
• •Use references/model-directory-spec.md and references/model-update-policy.md.
• •Validate local models using scripts/check_models.py.

1. •Implementation (Phase-by-Phase)

• •Execute the next step only after verification passes.
• •Record results and failures in project files.

1. •GUI Quality Requirements

• •Define a clear UI style direction.
• •Apply design-system conventions and responsive layout.
• •Use references/gui-quality-checklist.md.
• •If needed, delegate UI design to the frontend-design skill.

1. •API Contract (Required)

• •Define a common API template before implementation.
• •Cover auth, input schema, output schema, errors, and versioning.
• •Use references/api-contract-template.md.
• •Include at least one request/response call example per endpoint.

1. •Delivery & Portability

• •Ensure project can be packaged and shared.
• •Provide a clean startup guide and offline model usage guide.
• •Use references/packaging-checklist.md.
• •Use references/deployment-packaging-template.md for packaging plans (desktop/Docker/zip).

Skill Coordination

• •Use fastapi for Python API architecture decisions.
• •Use frontend-design for high-quality GUI direction.
• •Use planning-with-files to store plan artifacts in repo.
• •Use verification-before-completion to enforce test gates.
• •Use architecture-patterns for decoupled structure.

Required Artifacts

• •README.md with setup, run, and offline model instructions.
• •Environment report (hardware + toolchain).
• •Plan file with verification checklist.
• •Research source log with dates.
• •Model manifest(s) with source, version, checksum, date.
• •Conflict matrix and resolution decisions.
• •API contract document.

Validation Discipline

• •Treat every step as incomplete until its verification passes.
• •If verification fails, record the failure and fix before proceeding.