development-flow-loop
Structured software delivery lifecycle for coding tasks that should move from planning to production with repeated quality gates. Use when building features, refactors, integrations, automations, or s
SKILL.mdresearch/sphere-grid/skills/development-flow-loop/SKILL.md
GitHubDevelopment Flow Loop
Overview
Follow this exact execution loop for development work. Advance only when each stage gate is satisfied.
Lifecycle
- Plan
- Define objective, constraints, dependencies, risks, and success criteria.
- Output: concise plan with explicit deliverables.
- Architect
- Choose structure, interfaces, data flow, and failure boundaries.
- Output: implementation shape and file ownership.
- Evaluate
- Critique the design before coding: complexity, maintainability, testability, cost, and safety.
- Output: go/no-go with concrete adjustments.
- Implement
- Ship the minimum correct version first.
- Keep changes scoped; avoid unrelated edits.
- Test
- Run targeted tests first, then broader checks if needed.
- Output: pass/fail evidence.
- Fix
- Resolve failures from step 5 with minimal blast radius.
- Test
- Re-run relevant tests until stable.
- Improve
- Upgrade quality after stability: readability, performance, docs, ergonomics, observability.
- Test
- Validate improvements did not regress behavior.
- Fix or Launch Decision
- If red or risky: fix and return to Test.
- If green and acceptable risk: prepare launch.
- Launch
- Publish/deploy/release with verification and rollback awareness.
- Output: release confirmation and post-launch check result.
Operating Rules
- Prefer small increments and frequent validation over large untested changes.
- Keep a clear audit trail: what changed, why, what proved it works.
- If blocked, reroute quickly, then merge best parts back into one path.
- Treat this flow as iterative: repeat
Fix -> Test -> Improve -> Testas needed.