SuperAPI AI Directory

full-stack-master

Testing & quality

View repository

Updated 10-15 20:10

About this skill

Full Stack Master Workflow Skill

I. Capability Positioning (Capability)

  • Workflow Auto-Orchestration: Connects the full chain of requirements, design, development, testing, quality, documentation, commit, review, and release.
  • Git Worktree Orchestration: Automatically selects physically isolated working directories based on task dimensions (dev/test/fix/docs) to accelerate parallel development.
  • Skill Aggregation: Integrates all core atomic skills (Todo Manager, Technical Planning, Backend Logic Expert, Vue Frontend Expert, Database Expert, Code Quality Auditor, Test Engineer, UI Validator, Conventional Committer, DevOps Specialist).
  • Reusable and Extensible: Can incorporate new scenarios (such as database migration, API changes, operational releases, etc.) and supports switching between multiple projects.
  • Phased Handoff/Dispatch: Can manually or via script assign phase tasks to the corresponding skills or specialized agents.

II. Mandatory Documentation (Mandatory Documentation)

Before performing any write operation or decision, you must ensure that you have read and understood the latest content of the following documents:

III. Shared Execution Principles (Shared Execution Principles)

  • This section only retains the execution summary from the full-stack orchestration perspective and no longer copies the complete project-level rules; the authoritative sources are respectively AI Collaboration Standards、Development Standards and Testing Standards .
  • Orchestration follows a four-step judgment order by default: first expose assumptions, then choose the minimal solution, then limit the scope of changes, and finally use minimal sufficient verification to decide whether to continue expanding.

IV. Standard PDTFC+ 2.1 Workflow (Standard Workflow)

  1. P (Plan) - Requirements Analysis and Planning

    1. Read Documents: Confirm todo.md、roadmap.md, the current acceptance criteria, and necessary standards.
    2. Scope Check: Determine whether the matter belongs to the current todo, the current acceptance criteria, or the current phase plan; if not, first follow planning.md to complete queue-jumping or deferral triage.
    3. Solution Design: Output the list of affected files, the verification matrix, and the phase handoff order; when needed, prioritize parallel splitting of read-only retrieval, audit evidence collection, and document comparison.
    4. Task Placement: Only enter Do for matters permitted for execution; if still in the assessment stage, full implementation must not be expanded in advance.
    • Skill:requirement-analyst、todo-manager、technical-planning、context-analyzer
  2. D (Do) - Development Implementation

    1. Single Owner: The same matter retains only one implementation owner at any point in time; by default it is handled by full-stack-master Overall coordination; only after boundaries are stable should frontend and backend specialized roles be assigned.
    2. Core Implementation: Follow development standards; if persistence is involved, handle the data model or entity first, then implement business logic and UI.
    3. Infrastructure Exception: When Docker, CI/CD, deployment, build environment, or operations scripts are involved, the full-stack lead remains responsible for unified orchestration, and reuse as needed devops-specialist。
    4. Release Script Entry: For unified pre-release validation, prioritize using scripts/release/pre-release-check.mjs; when Review Gate evidence is involved, prioritize using scripts/review-gate/generate-evidence.mjs to generate the initial record, then proceed to manual audit or supplementary evidence.
    5. Scope Gate: If new optimization points or non-blocking items are discovered during development, you must return to the P phase for re-triage, and must not silently expand the current task.
    • Skill:database-expert、backend-logic-expert、vue-frontend-expert、nuxt-code-editor、devops-specialist
  3. A (Audit) - Audit Release

    1. Review Gate: All changes to code, documentation, configuration, scripts, and governance definitions must enter code-quality-auditor audit.
    2. Security and Consistency: Simultaneously check security boundaries, Todo acceptance points, the validation matrix, the evidence chain, and residual risks.
    3. Return Strategy: If the audit fails or new issues beyond the current scope are found, return to D or flow back to P, rather than proceeding to subsequent phases with blockers.
    • Skill:code-quality-auditor、security-guardian
  4. V (Validate) - Browser and Flow Validation

    1. UI Validation: When page rendering, interaction flows, or visual behavior are involved, use ui-validator to complete browser-side validation.
    2. Skip Rule: If the current item has no UI-facing or flow-layer impact, you must explicitly record why the V phase can be skipped, rather than omitting it by default.
    • Skill:ui-validator
  5. T (Test) - Testing and Regression

    1. Read Testing Standards First: Before executing tests, you must read testing standards。
    2. Choose Validation by Risk: Based on the change type and budget, choose targeted tests, full tests, coverage,verify or performance validation; do not apply a one-size-fits-all full run.
    3. Test Flowback: If testing introduces new code changes or exposes blockers, you must return to D and go through the A phase audit again.
    • Skill:test-engineer、code-quality-auditor
  6. F (Finish) - Documentation and Closure

    1. Documentation Sync: Update todo.md status, and sync README, Guide, Standards, Design, and Plan documents as needed.
    2. Evidence Closure: At release or governance closure, prioritize using scripts/review-gate/generate-evidence.mjs to refresh the Review Gate evidence file, then summarize validation records, Review Gate conclusions, uncovered boundaries, and follow-up rerun plans.
    3. Commit Notes:Commit is not an independent main phase by default for this skill; only when the user explicitly requests a commit, or the current flow is explicitly assigned as a commit flow, should it be called after F-phase closure conventional-committer。
    • Skill:documentation-specialist、todo-manager、conventional-committer

V. Intent Extraction Methodology

  1. Progressive Refinement: First lock down the overall structure and goals, then drill into specific implementation details.
  2. Single-point breakthrough: Ask only one question at a time, and wait for the user to answer before proceeding to the next follow-up question.
  3. Loop verification: When the user's answer is unclear, try rephrasing to confirm.
  4. Intent extraction: Analyze the why behind what the user wants and provide better professional advice.

VI. Skill References (Each Sub-Skill Reference)

VII. Authoring Rules

  1. Imperative & Structured

    • Use verb + objective to describe the usage section of each step/each skill in a standardized way.
    • Long-winded filler and process-introduction-style text are prohibited.
  2. Clear inputs and outputs

    • Each step must specify the input dependencies and output artifacts of this stage (such as file paths, document links).
    • Example: Input: docs/plan/, Output: docs/design/xx.md.
  3. Chainable composition

    • Each step's skill should be allowed to run independently, or as a local fragment of the global master call chain.
    • Some skills support multi-role collaboration (such as testing and documentation running in parallel).
  4. Security checks and general exception handling

    • Forcefully insert quality gates such as typecheck and lint; entering the commit/release stage before detection is prohibited.
    • Clearly define security levels and data protection points.
    • Enforce a 'plan first, implement later' gate for items newly added mid-iteration; expanding scope while working is prohibited.
  5. Internationalization and documentation first

    • The creation of all workflows/skills should be compatible with i18n and standard documentation synchronization actions by default.

VIII. Template Usage (Usage Example)

workflow:
  - step: "需求分析"        # context-analyzer, documentation-specialist
  - step: "功能开发"        # nuxt-code-editor
  - step: "UI 验证"         # ui-validator
  - step: "质量检测"        # code-quality-auditor, security-guardian
  - step: "功能提交"        # conventional-committer
  - step: "测试补充"        # test-engineer
  - step: "测试提交"        # conventional-committer