Execute hierarchical engineering plans phase-by-phase with TDD and stateful memory.
npx skills add https://github.com/luyzkk/Anti-Vibe-Coding --skill execute-plan-luyzkk
Please help me install this Agent Skill. Skill: execute-plan Source: https://github.com/luyzkk/Anti-Vibe-Coding/tree/main/skills/execute-plan Command: npx skills add https://github.com/luyzkk/Anti-Vibe-Coding --skill execute-plan-luyzkk
execute-plan turns a hierarchical engineering plan into reliable, verifiable progress by orchestrating isolated subagents per phase and keeping TDD and state/memory consistent.
Run execute-plan for your feature plan by choosing (or providing) the PLAN.md/active plan folder and confirming execution when prompted.
High-intent search queries and answers about installing and using this skill.
You execute hierarchical plans phase-by-phase by orchestrating isolated subagents that enforce RED-first test creation and GREEN-only minimal implementation while keeping context clean between cycles. The process updates stateful memory and validates completion criteria per phase.
Orchestrating subagents for isolated TDD cycles involves dispatching phase-specific tasks to clean contexts, enforcing deterministic RED/GREEN test constraints, and capturing live progress into STATE.md and MEMORY.md artifacts for reliable execution.
Stateful memory tracking works by updating STATE.md and appending rich implementation notes like DI-*, BUG-*, and DEV-* into specific MEMORY.md sections as each phase completes, ensuring consistent context retrieval across isolated subagents.
Yes, you can use this approach with a legacy flat PLAN.md format. The execution logic supports both the v2 folder-based layout with waves/tasks and the v1 flat plan format, requiring interactive control and post-phase validation for both.
You emit a machine-readable completion signal and optionally move completed plans to a terminal directory after validating all phase completion criteria. This deterministic orchestration output supports downstream automation workflows.
Isolated agent context matters because it enforces strict TDD boundaries by preventing context bleed between RED and GREEN phases. This deterministic isolation ensures minimal implementation and reliable state updates across complex hierarchical plans.
LinkDryPackage validated Phase 1 lines into Phase 2 publication artifacts with evidence references.
trpc-groupWrite a deterministic OK marker to out/ok.txt via a shell script.
anshul-garg27Automate Red-Green-Refactor TDD cycles with subagent isolation across JavaScript, TypeScript, Python, Go, and Java.
Pear-CommerceExecute phase implementation plans in dependency-aware parallel waves via subagents.
