skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
robsonrung/rar-skills85 installs

coding-design-plan

Plan the implementation shape of one approved coding task. Use when a module, data, API, component, or deployment boundary still needs a design decision; use to-tasks for task decomposition.

How do I install this agent skill?

npx skills add https://github.com/robsonrung/rar-skills --skill coding-design-plan
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    The skill provides a structured workflow for planning code implementation tasks. It contains no executable code, network operations, or malicious patterns.

  • Socketpass

    No alerts

  • Snykpass

    Risk: LOW · No issues

What does this agent skill do?

Coding Design Plan

Turn one approved task into the smallest coherent shape an implementer can build without making a new product or architecture decision. Choices settled in the PRD or task are already decided; build on them.

Workflow

  1. Read the task, its acceptance contract, and the project's conventions. Inspect the relevant code, tests, and current behavior.
  2. State the design problem in one sentence: the behavior that changes and its owner.
  3. Map only the affected boundary: module, component, service, data store, public contract, or deployment path. Use change ownership to place each responsibility where its next change belongs.
  4. Choose the smallest coherent shape that fits the existing code. When a real alternative exists, state the one trade-off that decides it. When the code makes one local move clear, take it.
  5. Add a check only when its trigger applies:
    • The task crosses a nonlocal boundary and has no lens conclusions yet: run design-gate.
    • The task already carries lens conclusions: apply them as constraints.
    • One uncertainty only running code can answer: use to-prototype. Prototype code never graduates.
  6. Name a test or direct check for every changed behavior. For a nonbehavior change, say why a test does not help and name the replacement check.
  7. Put adjacent work you discover in deferred. Widening scope needs a new task or the user's instruction.

Use a safe default for nonmaterial details and state the assumption. Ask the user only when a fact you cannot inspect would change the shape, public behavior, risk, scope, or cost.

Output

The plan is done when every field is filled and every changed behavior has its check.

  1. design_problem: one sentence.
  2. chosen_shape: the smallest coherent shape.
  3. boundaries: affected owners, modules, contracts, or runtime paths.
  4. constraints: task decisions and lens conclusions that apply.
  5. verification: the command, test, or direct check for each changed behavior.
  6. test_scenarios: input, action, and expected outcome per changed behavior, or Test expectation: none with its reason and replacement check.
  7. deferred: adjacent work kept out of this task.
  8. blocker_or_assumption: only an unresolved fact that could change the plan.

Write an ADR only when all three or no ADR holds: the decision is hard to reverse, surprising without context, and a real trade-off. Record on settle, not at the end of implementation.

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/robsonrung/rar-skills/coding-design-plan">View coding-design-plan on skillZs</a>