cube

Reconstruct and explain the business process implemented or affected by a feature, use case, module, endpoint, event, command, behavior, or problem, using system artifacts as evidence rather than analyzing code in isolation.

Audit selected code for material security, correctness, reliability, performance, and maintainability risks; report evidence-based findings with stable IDs, then implement only findings explicitly selected by the user.

Deliver the completed task result as one user-accessible Markdown document without changing its meaning, scope, workflow, or domain-specific output contract. Use only when explicitly invoked.

Summarize what has already been established in a task, including achieved outcomes, decisions, open points, and next steps, without restarting the analysis. Use only when explicitly invoked.

Split uncommitted repository changes into logical commits and create them using Conventional Commits.

Discover and resolve unresolved product, project, design, and architectural decisions through collaborative, dependency-aware interview rounds; do not use it for discoverable facts or immaterial execution details.

Turn established problem analysis, requirements, and decisions into a concrete step-by-step implementation plan with exact changes, dependencies, acceptance criteria, and verification.

Control response depth by producing the most complete useful answer with relevant explanation, evidence, implications, and limitations without unnecessary verbosity.

Drive a problem, idea, goal, requirement, symptom, or change proposal toward the strongest justified solution using focused evidence, explicit decisions, clarification, and solution comparison.

Decision principles for writing and refactoring source code and tests, applicable regardless of language, framework, or libraries. Prefer clear responsibilities, the simplest correct structure, meaningful domain boundaries, observable contracts, real dependencies, and justified abstractions.

Analyze selected code, propose only justified behavior-preserving refactorings with stable T1, T2, T3, and subsequent identifiers, then implement and verify only the scope selected by the user.

Control response presentation by producing the minimum sufficient answer while preserving required information, evidence, caveats, and output contracts.

Break non-trivial tasks into the minimum set of coherent, dependent, and verifiable execution stages while preserving the consuming skill's domain workflow.

Always use this skill before retrieving information from a repository, dataset, database export, API response, log, document set, test output, diff, or other structured or unstructured data source. It defines the minimum evidence and retrieval scope needed to make a correct decision while controlling context size. Use it for reading, searching, investigating, editing, reviewing, debugging, planning, and validating work that depends on external or workspace data. Do not use it when no data retrieval is needed, such as pure drafting from user-provided facts.