菜单
Manages workflow state for multi-agent design workflows. Tracks current phase, iteration count, document status, and agent activity. Use when orchestrating multi-phase workflows that need persistent state between agent invocations, or when checking workflow progress.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
AWS infrastructure reference knowledge for Terraform deployment, AWS services configuration, AMI building with GPU/NVIDIA drivers, and deployment troubleshooting. Provides production-ready reference docs for ALB, ASG, Lambda, CloudWatch, Image Builder, and Terraform IaC patterns.
Detects infinite loops and stale iterations in design-review cycles. Compares Critic feedback across iterations to find repeated issues. Use when managing iterative review workflows to prevent wasted cycles on unresolvable issues.
Structured markdown logging for multi-agent design workflows. Appends formatted events to workflow_log.md with timestamps, phase info, and agent activity. Use when recording workflow progress, phase transitions, Critic verdicts, or final summaries.
| name | workflow-state-manager |
| description | Manages workflow state for multi-agent design workflows. Tracks current phase, iteration count, document status, and agent activity. Use when orchestrating multi-phase workflows that need persistent state between agent invocations, or when checking workflow progress. |
Run the state tracker script to create a new state file:
python3 .github/skills/workflow-state-manager/scripts/state-tracker.py init --folder generated_docs_[TIMESTAMP]
This creates generated_docs_[TIMESTAMP]/workflow_state.json with initial state.
python3 .github/skills/workflow-state-manager/scripts/state-tracker.py set-phase --folder generated_docs_[TIMESTAMP] --phase [research|design|review|delivery]
python3 .github/skills/workflow-state-manager/scripts/state-tracker.py increment-iteration --folder generated_docs_[TIMESTAMP]
python3 .github/skills/workflow-state-manager/scripts/state-tracker.py set-verdict --folder generated_docs_[TIMESTAMP] --verdict [APPROVED|CONDITIONAL|REJECTED]
python3 .github/skills/workflow-state-manager/scripts/state-tracker.py status --folder generated_docs_[TIMESTAMP]
The script manages workflow_state.json:
{
"project": "...",
"started": "2026-02-10T17:28:21",
"phase": "review",
"iteration": 2,
"max_iterations": 5,
"verdict": "CONDITIONAL",
"documents": {
"solution_design": "draft"
},
"research_files": [],
"history": []
}