← ClaudeAtlas

implementation-planlisted

Creates or repairs Slice-first planned-feature artifacts for approved changes. Use for initial planning, same-requirement plan repair during auto-resolve implementation, package breakdowns, or task artifacts. Do not use to perform coding, code review, audit, or status.
leninkhaidem/super-developer · ★ 0 · Code & Development · score 62
Install: claude install-skill leninkhaidem/super-developer
# Implementation Plan Orchestrate a fresh Slice-first planned-feature set under the selected artifact root: `SPEC.md`, lightweight `tasks.json`, package Markdown, declared proof/report paths, and proof placeholders when dispatch is next. “Fresh” describes the artifact set, not whether the target system is new. Source inspection and helpers use the separate code root. This skill is the `implementation-plan` **orchestrator**, not the planner worker. Preserve approved requirements, roots, decisions, and workflow state here; delegate artifact writing with `references/planner-agent-contract.md`. Mode is `initial` by default. Use `implementation-continuation` only when `implement` supplies approved requirements, roots/ref/slug, current artifacts, the Execution Contract, a stage-sourced plan-defect packet, and accepted empirical reports or explicit `none`. ## Always - Plan from approved requirements, safe Conceptualize/diagnosis material, verified repository/official evidence, and, when empirical evidence applies, accepted bounded `empirical-spike` reports only. Preserve planned-hotfix context without inventing an artifact field or feature ref. - Prefer static/official evidence. Inventory a bounded set of distinct material empirical questions tied to the current approved decisions; routine work, cost alone, and statically resolved questions do not trigger a spike. - Preserve planning context while evidence runs. Give each falsifiable question one stable logical-question ID