Execute implementation using Ralph Wiggum iterative loops on specs
$ARGUMENTS
You MUST consider the user input before proceeding (if not empty).
This skill launches the Ralph Wiggum implementation loop to process specifications autonomously. The agent iterates until acceptance criteria and Completion Signal requirements pass.
At least one spec: Specs must exist in specs/ with Completion Signal sections
Context files:
.specify/memory/constitution.md — Project principlesAGENTS.md — Development guidelinesRALPH_PROMPT.md — Master prompt (optional, can use inline)If $ARGUMENTS specifies a single spec (e.g., "001-user-auth"):
specs/$ARGUMENTS/spec.md.specify/memory/constitution.md for project principlesAGENTS.md for development guidelines<promise>DONE</promise> when ALL checks passIf no specific spec provided, run the master loop:
.specify/memory/constitution.md for project principlesAGENTS.md for development guidelinesspecs/ folder in numerical orderspecs/{spec-name}/spec.md<promise>ALL_DONE</promise> when all specs completeYou can also run the Ralph loop via shell:
./scripts/ralph-loop-codex.sh
Or for a specific spec:
./scripts/ralph-loop-codex.sh 001-user-auth
PyTorch深度学习模式与最佳实践,用于构建稳健、高效且可复现的训练流程、模型架构和数据加载。