Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add dhslegen/digital-delivery-team --skill relaygit clone --depth 1 https://github.com/dhslegen/digital-delivery-teamWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/dhslegen/digital-delivery-team/relay)<a href="https://agentmods.dev/skills/dhslegen/digital-delivery-team/relay"><img src="https://agentmods.dev/badge/skills/dhslegen/digital-delivery-team/relay/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/dhslegen/digital-delivery-team/relay"><img src="https://agentmods.dev/badge/skills/dhslegen/digital-delivery-team/relay.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00111 | $0.02377 |
| Opus 5 | $0.00056 | $0.01189 |
| Sonnet 5 | $0.00022 | $0.00475 |
| Haiku 4.5 | $0.00011 | $0.00238 |
Grade A, and why
relay scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 203 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DDT Relay · 跨会话接力 skill
DDT 13 段式接力 prompt 范式：覆盖 What WORKED / What Did NOT Work / Decisions / Next Step 等关键信息，叠加 DDT 项目特定背景注入（progress.json / tech-stack.json / 关键产物路径自动捕获）。
何时触发
只要用户表达"换会话续作"的意图，立即启动本 skill：
- 显式触发：
/digital-delivery-team:relay或/relay - 关键词触发："接力"、"换会话"、"换 AI"、"导出进度"、"context 不够"、"token 不够"、"备忘录"、"传家宝"、"会话切换"、"接力提示词"
- 隐式信号：会话已运行超过 4 小时 / 工具调用次数 ≥ 200 / 连续 3 次提示 compact
工作流
Step 1: 收集材料（让 main thread 自己做，不开 subagent）
读取以下来源（按优先级）：
.ddt/progress.json— 阶段进度状态机.ddt/tech-stack.json— 技术栈选型.ddt/project-id— 项目 IDproject-brief.md— 原始需求docs/*.md— 已落盘的产物（PRD / WBS / arch / blockers / review-report ...）git log --oneline -30— 近 30 次提交（已确认有效的工作）git status --short— 未提交的改动（在制品）git diff --stat HEAD~10— 最近 10 个 commit 的变更摘要- 当前会话的对话历史 — 抽取本次会话内 LLM 自己识别的关键决策、失败尝试、未尝试方向
Step 2: 调用 build-relay-prompt.mjs 生成结构化 prompt 文件
node "$DDT_PLUGIN_ROOT/bin/build-relay-prompt.mjs" --out .ddt/relay-$(date +%Y%m%d-%H%M%S).md
脚本自动注入：
- 项目 ID / 当前 phase / 已完成 phase
- 技术栈摘要
- 关键产物绝对路径
- git 历史摘要
Step 3: LLM 在生成的 prompt 文件基础上补充会话内容
LLM 必须填写的 9 个段落（13 段中其余 4 段由脚本自动填）：
- What We Are Building — 1-3 段：项目目标 + 为什么 + 系统位置（参考 brief + PRD 第 1 段）
- What WORKED (with evidence) — 已验证有效的产出（每条带具体证据：test 通过 / lint 0 错 / git commit / 用户确认）
- What Did NOT Work (and why) — 失败的尝试 + 精确原因（防止下次重试同一坑，最关键段）
- What Has NOT Been Tried Yet — 待尝试的方向（含原因）
- Current State of Files — 表格：路径 / status (✅ / 🟡 / ❌ / ⏸) / Notes
- Decisions Made — 架构 / 技术选型决策 + 理由
- Blockers & Open Questions — 未解决的阻塞 + 待回答的问题
- Exact Next Step — 下次会话的第一步（精确到无需思考）
- Environment & Setup Notes（可选） — 启动命令 / 环境变量 / 依赖
禁忌：
- 不要把"未确认有效"的工作放进 What WORKED（移到 What Has NOT Been Tried Yet）
- 不要省略 What Did NOT Work（即使写"无失败尝试"也比省略好）
- 不要在 Exact Next Step 里写"看情况"、"视需要"——必须精确
Step 4: 输出完整 prompt 到屏幕
把整个 prompt 用 text 代码块包裹输出到屏幕，让用户能一键 copy。
显著标注："请整段复制下面的 prompt 到下一个会话开头"。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 203 lines · 111 tokens per session scan A 43634d64a323
relay is a skill published in the GitHub repository dhslegen/digital-delivery-team (1 stars, last pushed 3mo ago), licensed MIT. It adds 111 tokens to every session and 2,377 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
team-memory-protocol
A shared-memory process for teams of coding agents working on the same build. It stores project knowledge, shared context, and task progress in separate files or repositories.
knowledge-wiki
A tool that turns useful lessons from coding sessions into general articles for a team wiki. A wiki is a collection of linked reference pages, and this one can be kept locally or shared through a Git repository.
using-compounded
Use this skill at the start of any session where compounded is installed, or whenever the user mentions earning trust, verifying skills, the trust ladder, compounded, the trust gradient, autonomous skills, or asks how compounded works. Explains the four trust states (.proposed, .verified, .trusted, .autonomous) and…
design-system-reference
Style guides and implementation rules for frontend design. Works with design-discovery agent which handles context gathering and VS-based style recommendations. Contains detailed style guides, anti-patterns, and implementation checklists.
dream
Prunes and consolidates the project's auto-memory to keep it minimal — deletes stale, wrong, or redundant memories, merges overlapping ones into existing files, and trims the index. Strong bias against growth: never creates new memory files and never stores new facts. Verifies staleness against the current codebase…
magic-compact
Compact this Claude Code session.