Agent Skills
› AgriciDaniel/claude-ads
› ads-plan
ads-plan
GitHub用于创建专业付费广告策略的 Skill,涵盖目标设定、渠道评估、预算规划、实验设计及发布流程。
Trigger Scenarios
制定广告投放计划
设计媒体购买策略
规划 PPC 或社交广告架构
Install
npx skills add AgriciDaniel/claude-ads --skill ads-plan -g -y
SKILL.md
Frontmatter
{
"name": "ads-plan",
"description": "Create a professional paid-advertising strategy covering objectives, economics, platform selection, campaign architecture, audiences, budget, creative, measurement, experiments, governance, rollout, and reporting. Use for ad plan, media plan, PPC strategy, paid-social strategy, campaign architecture, advertising roadmap, or channel planning."
}
Paid Media Plan
- Read the setup profile, business economics, evidence, current account state, and relevant industry template as optional input.
- Define objective, customer, offer, conversion, value, constraints, geography, regulated category, time horizon, and success criteria.
- Evaluate channel roles and exclusions from first principles; do not require every
supported platform. For a channel outside the twelve-platform contract, load
ads/references/additional-platforms.md(resolve it from the plugin root, the directory that holdsads/andskills/, or from the skills directory in a standalone install) and return a research lead unless its current buying, eligibility, measurement, and creative evidence is present. - Before proposing Meta budgets, learning-phase expectations, bidding,
consolidation, or performance forecasts, collect the account, Pixel, and
conversion cold-start dimensions per the contract in
skills/ads-meta/SKILL.md. When any dimension is cold orunknown, plan for measurement validation, explicit creative hypotheses, staged reversible tests, and confidence labels; do not apply mature-account benchmarks to missing history. - Specify campaign architecture, audience strategy, creative system, budget and pacing, measurement, experiments, policy controls, and operating cadence.
- Phase prerequisites before launch, learning, optimization, and scale.
- Assign every action an owner, timing, dependency, guardrail, evidence, success measure, and rollback or exit condition.
- Return canonical JSON and render the requested human plan.
A plan is advisory. It becomes an account change only through launch or optimize mutation gates.
Version History
-
fc3908e
Current 2026-10-19 09:33
修复了插件根目录共享引用的解析可靠性问题,增加了安装器引用固定功能,并优化了多平台技能的参考路径提示。
-
ac21644
2026-09-22 14:42
新增Meta冷启动合同检查,在制定预算和预测前强制收集账户、Pixel及转化历史数据,避免对缺失历史使用成熟期基准。
- 669c760 2026-07-25 07:46


