Tasks & synthesis · Plan

Evolution & Improvement

Produce an optimal, realistic, repo-grounded improvement path for 1–3 iterations. The plan should increase correctness, product quality, regression resistance, delivery speed, or hot-path performance without inventing work just to fill a template.

Vibe Coding 1.3.0 · vibe-task · English technical instructions

← Back to search

Use this workflow

Use this workflow in Codex or Claude Code with the free Vibe Coding plugin. Choose your assistant and prompt language, then add your task details after the prompt.

Operation

Analyze and return an actionable brief in the current conversation. Planning does not create a new assistant task or edit product code. Carry the full requested objective into the plan; separate independent work without silently discarding it.

Goal

Produce an optimal, realistic, repo-grounded improvement path for 1–3 iterations. The plan should increase correctness, product quality, regression resistance, delivery speed, or hot-path performance without inventing work just to fill a template.

Audit focus

Map the current project from repository evidence:

  • key domains/modules, source-of-truth layers, public surfaces, critical user/admin/system flows;
  • business invariants, state transitions, tenant/role/plan rules, integration contracts, migrations, jobs, and tests;
  • UX friction in core flows: unclear navigation, missing states, unsafe copy, form/table/destructive/action ambiguity;
  • regression risks: duplicated logic, weak tests, generated/docs drift, brittle boundaries;
  • performance and operability issues only when a reachable path is visible.

Planning rules

  • Every major recommendation and task spec must cite path:line[-line] evidence plus symbol when possible.
  • Distinguish direct evidence from inference. Do not present inferred traffic, business priority, or product intent as repo fact.
  • Do not elevate theoretical risks into plan items without a concrete reachable path, contract exposure, or observable inconsistency.
  • Collapse symptoms with the same behavior owner and authoritative fix unless separate rollout units are clearly required.
  • Use exactly one solution direction per task; no alternatives.
  • Optimize for maximum impact per unit of risk/cost, accounting for dependencies and incremental delivery.
  • Omit empty phases. If the audit finds no actionable items, state what was checked and say no actionable evolution items were found by static audit.

Artifact structure

evolution_plan.md should be practical, compact, and scannable. Include sections that fit the repo:

  • Baseline: architecture map, critical flows, current pain points, constraints.
  • North Star: UX/domain/engineering outcomes, using measurable or proxy metrics when available.
  • Roadmap: 1–3 phases with goal, scope, deliverables, dependencies, validation, and rollback/migration notes when relevant.
  • Task Specs: atomic roadmap tasks with ID, priority, theme, problem, evidence, root cause, impact, fix direction, steps, acceptance criteria, and validation commands when they exist in the repo.
  • Explicit Non-Goals: what is intentionally excluded to avoid scope creep.

Keep task specs implementation-ready, but these are roadmap specs, not task briefs output. If a future step needs task briefs, use Synthesis or Task Shaping.

Workflow ID: evolution-improvement · View the versioned source · Shared workflow and authority rules