Overseas Ad Attribution Link Skill
Operating Mode
Use this skill for two output types:
- Technical solution design for overseas ad attribution link capabilities.
- Q&A and troubleshooting for attribution, parameters, SDK timing, PushAPI, landing pages, deep links, event callbacks, reporting controls, and post-install routing behavior.
Identify the mode first. Use solution-design mode when the user asks for a technical solution, integration plan, product/technical review, feature design, or architecture. Use Q&A mode when the user asks why a chain behaves a certain way, how to judge a source, where a parameter comes from, how a link stage works, or how to troubleshoot a metric or behavior symptom.
Required Reference Loading
Load only the references needed for the request:
- For technical solution or functional architecture: read
references/technical-solution-template.md. - For attribution rules, data-source priority, interface timing, parameter mapping, or event callbacks: read
references/attribution-knowledge.md. - For usage questions, diagnosis, or explanation: read
references/qa-playbook.md.
Do not load or cite customer-specific source documents unless the user explicitly provides them for the current task.
Solution-Design Workflow
When drafting an overseas ad attribution link solution:
- Clarify the scenario: deep link, landing page, SDK attribution, PushAPI, Google Install Referrer, or mixed chain.
- Identify the business goal: attribution completion, ad hierarchy backfill, event callback, post-install routing to promoted content, model-learning signal, reporting control, or metric reconciliation.
- Define actors and modules: media platform, ad-link generator, landing-page smart script, app client, SDK/MMP, backend attribution service, reporting service, admin configuration, and data warehouse.
- Specify data sources and priority. Document each source's timing, reliability, matching key, fallback behavior, and allowed use.
- Specify client timing: cold start, first install open, privacy consent, SDK callback timing, fallback timeout, short polling, hot start, and session-level limits.
- Specify backend behavior: matching, deduplication, spread mapping, ad hierarchy backfill, event eligibility, and callback routing.
- Specify user-facing behavior: auto-route, route prompt, route suppression when the user is already consuming content, or delay until attribution result arrives.
- Specify validation: cases by source/channel, parameter examples, expected logs, event callbacks, metric checks, gray release, and rollback switches.
Use references/technical-solution-template.md as the structure. Remove sections that are irrelevant to a narrow request.
Q&A Workflow
When answering usage questions:
- Restate the observed symptom or decision point.
- Name the relevant chain stage: ad click, landing page, install, first open, SDK callback, pre-attribution, post-attribution, ad-info request, event callback, or analytics reconciliation.
- Map it to known rules in
references/attribution-knowledge.md. - Explain the likely cause with source priority and timing, especially when PushAPI, Google Install Referrer, or SDK data arrives late.
- Give a verification checklist: logs, fields, request timing, user state, spread/ad hierarchy fields, event eligibility, and reporting switch.
- Separate confirmed facts from assumptions. If the current project has its own rules, ask the user for the relevant rule or document excerpt before overriding the generic pattern.
Domain Guardrails
- Distinguish attribution completion from post-install routing/playback success; a user can be attributed but not routed to promoted content.
- Distinguish pre-attribution and post-attribution. Pre-attribution affects immediate app behavior; post-attribution can complete attribution later but may require polling or routing compensation.
- Treat Google Install Referrer as first-install-open oriented. Do not assume it updates for every new ad click after install.
- Prefer explicit source-priority rules over "latest field wins" unless the user states a project-specific rule.
- When discussing event callback, separate media event mapping, eligibility rules, deduplication, and reporting switches.
- When a metric issue appears, inspect both attribution timing and app experience before concluding the attribution logic is wrong.
Sensitive-Context Guardrails
- Do not include company names, project names, document names, historical dates, personal names, local file paths, tokens, URLs, package names, campaign names, or raw log examples in generated reusable guidance.
- Public media/platform names such as Facebook, TikTok, Snapchat, and Google may be retained when relevant to solution design or troubleshooting.
- Internal company names, project names, document titles, campaign names, package names, and raw account identifiers must be removed or generalized.
- Use generic categories such as "media SDK", "MMP SDK", "social media channel", "ad network", "landing-page log", or "install-source API" when the source is internal, customer-specific, or not necessary to the answer.
- Use relative timing such as "initial request", "fallback timeout", and "short polling window"; avoid historical dates unless the user supplies dates for the current deliverable.