← ClaudeAtlas

right-size-engineeringlisted

Calibrate software engineering decisions across macro, meso, and micro perspectives so the result is complete without unsupported structure, detail, or scope. Use when requirements, architecture, design, implementation, migration, or verification may be misaligned; when local and structural approaches compete; when a decision introduces enduring commitments; or when over- or under-engineering is a material risk. Avoid for fully specified mechanical work, unconstrained exploration with no decision to make, or tasks owned by a more specific review method.
hudrazine/velkross · ★ 1 · AI & Automation · score 62
Install: claude install-skill hudrazine/velkross
# Right-Size Engineering Make the least speculative engineering commitment that completely supports the required outcome. Use macro, meso, and micro as relative perspectives on the current subject, not fixed development phases: - **Macro — purpose and system effect:** outcome, users, contracts, constraints, non-goals, and consequences - **Meso — structure and ownership:** responsibilities, boundaries, relationships, workflows, dependency direction, and propagation - **Micro — mechanism and evidence:** behavior, state, algorithms, failure handling, configuration, tests, and operational details A project, subsystem, module, API, or function can each be examined through all three perspectives. Do not treat a perspective as a measure of importance, change size, or permission to expand scope. ## Calibrate The Decision 1. Identify the required outcome, current decision, hard constraints, available evidence, and material uncertainty. 2. Start at the perspective where the decision appears to belong. 3. Inspect the nearest adjacent perspective when needed: - look upward to justify purpose, scope, or enduring structure - look downward to confirm feasibility, behavior, failure semantics, or verification 4. Place the decision at the boundary that owns its consequences. 5. Choose the least irreversible solution supported by present requirements and evidence. Do not perform a complete three-perspective analysis by default. Traverse only far enough to resolve material