sdlc-create-plan

Creates locked dependency-safe execution plans with task graphs for one feature.

7|5|Updated Jul 22, 2025
One-click install
npx skills add https://github.com/nebius/nebius-ps-services --skill sdlc-create-plan-nebius
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sdlc-create-plan
Source: https://github.com/nebius/nebius-ps-services/tree/main/skills/sdlc-create-plan
Command: npx skills add https://github.com/nebius/nebius-ps-services --skill sdlc-create-plan-nebius

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and assets (resource) components.

What problem does it solve? Turning a ready feature specification into a safe, parallelizable implementation plan is error-prone: tasks conflict over shared files, plans get edited after locking, and post-evaluation fixes risk rewriting completed work. This Skill produces a local, locked execution plan for exactly one feature and enforces append-only corrective plan versions after a classified evaluation failure. ## Core Features & Use Cases - Dependency-Safe Task Graphs: Expresses implementation as stable TASK-* records with dependencies, exact/prefix write claims, conflict domains, focused validation, done criteria, and stop conditions. - Vertical Slice Planning: Preserves one end-to-end slice across application layers instead of splitting work into broad layer-only steps. - Immutable Corrective Plans: After a proven post-evaluation defect, creates plan vN+1 that preserves every existing task definition byte-for-byte and appends diagnosis-bound corrective tasks, validated by scripts/corrective_plan.py. - Use Case: A coordinator reaches the planning phase for FEAT-001 with requirements and design docs ready; the Skill writes plans/FEAT-001.plan.v1.md plus a lock file, and later, after a failed evaluation, appends a validated corrective v2 plan bound to the original regression oracle. ## Quick Start Ask the Agentic SDLC coordinator to run the create-plan phase for the current ready feature so it writes and locks the local execution plan.

Frequently Asked Questions about sdlc-create-plan

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a locked execution plan for a feature?

Invoke the Skill through the verified Agentic SDLC coordinator with one ready FEAT-* and its REQ-* blocks. It writes plans/FEAT-*.plan.vN.md plus a .lock file containing TASK-* records, dependency waves, validation commands, and stop conditions.

How do I fix a failed evaluation without rewriting completed tasks?

Create an adjacent corrective plan vN+1 that preserves every existing task definition byte-for-byte and appends new diagnosis-bound TASK-* records. Run scripts/corrective_plan.py to validate append-only history, the completed task manifest digest, and the original regression oracle before dispatch.

Can I use sdlc-create-plan standalone outside the SDLC workflow?

No. The Skill requires verified workflow context from the active Agentic SDLC coordinator, including the selected project, active run, and phase. Without that evidence it returns without mutation, and it is hidden from the public command menu in Claude Code.

When should tasks be serialized instead of run in parallel?

Serialize tasks that share files, path prefixes, APIs, schemas, migrations, dependency manifests, shared abstractions, or exclusive test resources. Mark uncertain ownership as unknown rather than inferring parallel safety from missing data.

What are the limitations of this planning Skill?

It never commits plans, edits locked plans, writes code or tests, or expands feature scope. If design is not implementable it routes to sdlc-create-design, and if plan lineage cannot be preserved it stops with HUMAN_INPUT_REQUIRED.