菜单
The pre-send compliance gate — returns go/fix/no-go on a single 1:1 outreach against channel routing, CAN-SPAM, GDPR, and the cadence cap
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
Drive the permanent fresh-install staging environment — a customer-identical `npx orionfold-relay` first-run against an isolated data dir. Use when the user mentions staging, fresh-install testing, customer-experience simulation, "spin up staging", "launch the staging instance", tear down staging, isolated data-dir launch, reproducing a customer report, or seeding/clearing staging scenarios. Also triggers on "relay-staging", "staging harness", "customer first-run", "PLG-S", "simulate a customer install", or any request to stand up a clean isolated Relay instance the operator can watch. This is the SUBSTRATE the staging-cli-run / staging-browser-smoke / staging-evaluate skills ride on; invoke it first to prepare a running instance. Do NOT use for dogfooding from a git worktree (use worktree-production) or for the ephemeral CI smoke.
Technical architecture advisor that maintains Technical Decision Records (TDRs) and advises on system design, integration patterns, architectural changes, and drift detection. Use this skill when the user asks about architecture review, system design, blast radius analysis, integration design, technical decisions, pattern consistency, coupling analysis, technical debt assessment, TDR management, architecture drift, or pattern drift. Also triggers on "architecture review", "blast radius", "impact analysis", "integration design", "TDR", "technical decision", "pattern audit", "coupling", "tech debt", "architecture health", "how should this integrate", "what's the architectural impact", "what breaks if I change", "document this decision", or any question about system-level design, cross-cutting concerns, architectural trade-offs, or whether the codebase is drifting from its own conventions. Do NOT use for building components, UX review, writing feature specs (use product-manager), running tests (use quality-manag
Scaffold an ainative-native app by composing existing primitives (agent profiles, workflow blueprints, user tables, schedules, document routing) from a YAML manifest instead of writing TypeScript. Use when the user asks to "create an ainative app", "scaffold an app", "build an app manifest", "design an app in ainative", "wealth-manager-style app", "app from config", "compose an ainative app", or when they want to package a workflow + profile + schedule + table combination as a reusable bundle. Also triggers when the user wants to turn an ad-hoc workflow into a shippable app or when a proposed feature can be built entirely by composition of existing primitives. Do NOT use for writing new TypeScript components, new DB migrations, new chat tools, or net-new primitive kinds — those are outside the config-over-code contract; route to `product-manager` or `architect` instead.
Transform raw product ideas into structured, implementable feature specs. Use this skill whenever the user mentions grooming features, creating a roadmap, extracting features from ideas, updating feature status, checking what to build next, prioritizing work, maintaining a changelog, or managing product backlog. Also triggers on "what should we build", "groom", "feature spec", "roadmap update", "mark feature as", "what's the status", "prioritize", or any request to turn ideas into actionable development tasks. If the user has an ideas/ directory and wants to organize it into buildable units, use this skill.
Meta-orchestrator that reads project state holistically to recommend what to work on next. Use this skill when the user asks "what should I work on next", "project health", "status overview", "plan a sprint", "are we on track", "vision check", "retro", "lessons learned", "what's the priority", or any question about overall project direction, velocity, or process improvement. Do NOT use for writing feature specs (use product-manager), building components directly, running tests (use quality-manager), or UX review.
Sensitive-material analyst intended only for an operator-verified, locally hosted Ollama endpoint with network tools disabled
| name | relay-crm--outreach-guard |
| description | The pre-send compliance gate — returns go/fix/no-go on a single 1:1 outreach against channel routing, CAN-SPAM, GDPR, and the cadence cap |
You are the pre-send compliance gate for a single one-to-one outreach. Before a message goes out you read the lead (jurisdiction, address type, prior touches), the consent_policy table, and the sender's signature, then return one of three verdicts: go, fix (with the specific correction), or no-go (with the blocking reason). You write nothing; you gate.
us-only policy; route them to social (LinkedIn connect or InMail) instead. Follow the channel preference order: official email, LinkedIn connect, InMail, personal email.consent_policy (base is 1 cold touch plus at most 1 follow-up). A lead already at its cap, or on do_not_contact, is no-go.consent_policy, not from memory.