What problem does it solve? Managing a multi-phase software development lifecycle with an AI agent often loses state across context resets, mixes prompt intent with project specs, and retries failures without diagnosis. This Skill binds each run to an immutable prompt revision, checkpoints every state transition, and routes work through requirements, design, implementation, validation, and documentation phases with deterministic recovery. ## Core Features & Use Cases - Prompt-bound run coordination: Initialize a private prompt workspace and run objectives by exact prompt ref or file, with immutable revisions, FIFO queuing, and idempotent resume after context loss. - Checkpointed phase routing: Select exactly one next skill per iteration (requirements, design, steering, execution, commit, docs) based on checkpoints, fingerprints, and evidence rather than conversation memory. - Deterministic failure handling: Classify every gate failure through a repair-control contract, route ambiguous failures to troubleshooting, and enforce retry budgets and stop conditions. - Use Case: A developer initializes a workspace in a project folder, writes an Ask prompt describing a feature, and repeatedly runs the prompt; the coordinator drives requirements refinement, design validation, TDD execution, and documentation updates while surviving session restarts. ## Quick Start Ask the agent to run sdlc-start workspace init in your project folder, then run your prompt file with sdlc-start run to start the coordinated SDLC workflow.
