菜单
Use this skill when checking whether requirements, SDD documents, ADRs, code, tests, approvals, or deployment evidence are stale relative to their source references. Applies before implementation, review, approval, release, or when a user asks if docs/code/tests are up to date.
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。
Use this skill when checking whether the global Agentic SDLC setup is healthy across Codex, shared agents, Claude Code, OpenCode, GitHub Copilot, project-local skills, OpenCode /sdd routing, ADR indexes, SDD profile, and sync script configuration.
Use this skill when a user wants the full predictable AI engineering workflow, especially propose/apply/verify/archive, OpenSpec-like change lifecycle, Superpowers-style execution rhythm, Agent Skills-style discipline, or an end-to-end Agentic SDLC flow across Codex, Claude Code, OpenCode, or other IDE agents.
Generates a full design artifact set from a spec.md (primary source) and optional architecture.md (secondary source). Produces three outputs: architecture.md (system architecture), data-model.md (persistent/domain data model), and API_IMPLEMENTATION_GUIDE.md (API contract). Use this skill whenever a user provides a spec, architecture, or requirements document and asks for a design, detailed design, solution design, module design, or implementation design. Also trigger when the user says "convert architecture to design", "create design from architecture", "write a design document", "I need a design doc before we start coding", "break down this architecture for engineers", or "generate design artifacts". If the user uploads or pastes an architecture or spec document and wants to move toward implementation, this skill should always be used — even if they don't use the exact words "design document".
Use this skill when a task changes architecture, cross-project conventions, AI-agent working context, platform boundaries, integration rules, security posture, data ownership, or any decision that future maintainers and coding agents should remember. Also use when the user asks for context engineering, ADRs, decision records, durable project knowledge, or shared instructions for Codex, Claude Code, or OpenCode.
Use this skill when making Codex, Claude Code, OpenCode, GitHub Copilot, Gemini, or other coding agents share the same workflow skills. Creates or updates tool-specific instruction files, skill indexes, command shims, and global skill sync rules without duplicating workflow content.
Converts a detailed design document (design.md) into a structured implementation task breakdown (tasks.md) suitable for engineering execution, Jira decomposition, sprint planning, and handoff to coding agents. Use this skill whenever the user provides or references a design document and asks for implementation tasks, engineering breakdown, Jira tasks, sprint planning, delivery planning, or a work breakdown structure. Also trigger when the user says things like "convert this design into tasks", "break this down for engineering", "create an implementation plan from this design", "generate tasks from design", or "I want to hand this off to developers". Even if the user just uploads a design.md or pastes a design and says "what do I do next?" — use this skill.
基于 SOC 职业分类
| name | freshness-gate |
| description | Use this skill when checking whether requirements, SDD documents, ADRs, code, tests, approvals, or deployment evidence are stale relative to their source references. Applies before implementation, review, approval, release, or when a user asks if docs/code/tests are up to date. |
Use this skill to detect stale context before work proceeds.
Freshness is a first-class quality signal. A document, approval, test, or implementation can be correct for an old version and wrong for the current one.
Fresh - artifact is current relative to known upstream sources.Stale - upstream changed after this artifact was created, approved, or verified.Unknown - missing version evidence prevents a confident answer.Blocked - stale or unknown status must be resolved before proceeding.| Check | Stale When |
|---|---|
| Requirement to spec | requirement source changed after spec version |
| Spec to design | spec changed after design version |
| Design to code | implementation changed outside design scope or design changed after implementation |
| Test to requirement | tests do not reference latest requirement/spec IDs |
| Approval to document | document blob changed after approval |
| Deployment to code | deployed version does not match reviewed commit |
# Freshness Gate Report
## Verdict
Fresh | Stale | Unknown | Blocked
## Evidence
| Artifact | Version / Time | Source | Status |
|---|---|---|---|
## Findings
- [Finding]
## Minimal Repair Path
1. [Step]
## Open Risks
- [Risk]
Unknown; do not infer freshness.