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 xjli360/sealeap-amazon-ad-skills --skill sealeap-xiezhi-amazon-ai-product-research-governancegit clone --depth 1 https://github.com/xjli360/sealeap-amazon-ad-skillsWrote 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/xjli360/sealeap-amazon-ad-skills/sealeap-xiezhi-amazon-ai-product-research-governance)<a href="https://agentmods.dev/skills/xjli360/sealeap-amazon-ad-skills/sealeap-xiezhi-amazon-ai-product-research-governance"><img src="https://agentmods.dev/badge/skills/xjli360/sealeap-amazon-ad-skills/sealeap-xiezhi-amazon-ai-product-research-governance/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/xjli360/sealeap-amazon-ad-skills/sealeap-xiezhi-amazon-ai-product-research-governance"><img src="https://agentmods.dev/badge/skills/xjli360/sealeap-amazon-ad-skills/sealeap-xiezhi-amazon-ai-product-research-governance.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 | $0.00052 | $0.01222 |
| Opus 5 | $0.00026 | $0.00611 |
| Sonnet 5 | $0.00010 | $0.00244 |
| Haiku 4.5 | $0.00005 | $0.00122 |
Grade A, and why
sealeap-xiezhi-amazon-ai-product-research-governance 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 5d 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.
What it actually says
Amazon AI 选品判断治理
目标
让 AI 承担批量取数、整理和反证，而把竞品定义、需求真实性、差异化取舍与最终立项保留在人类决策门内。
适用任务
- 审计一个 AI 选品 Agent 是否只是加速了错误规则。
- 设计人机协作的选品流程与证据门槛。
- 复核 AI 给出的蓝海、竞争、需求或差异化结论。
开始前要拿到
- Agent 的提示词、工具清单、规则、阈值和样例输出。
- 候选产品事实、直接竞品定义和各数据字段的统计口径。
- 成功/失败案例、人工判断标准和风险偏好。
- 允许自动执行的动作与必须人工批准的动作。
缺少字段时列出证据缺口，并把相关结论标为 FACT、ESTIMATE、ASSUMPTION 或 UNKNOWN；不要补造数据。
不可妥协的边界
- 第三方数据均为估算或代理证据；Amazon 一方报告、后台实时字段和产品事实优先。
- 经验阈值只能作为可调起点，必须展示敏感性分析，不能写成 Amazon 官方规则。
- 不得捏造销量、搜索量、CPC、CVR、成本、认证、产品属性或消费者需求。
- 默认提供诊断或草案。写入前展示对象、旧值、新值、影响、停止线与回退，核对用户已有授权是否覆盖对象、动作与预算；范围已明确授权时继续执行并回读核验，只有未覆盖或扩大的范围才请求批准。
- 不输出或保存素材来源身份、账号、链接、作品编号、互动数据、原始话术或其他可反查来源的线索。
- 不得让 Agent 自动下单、发布或修改广告。
- 不得把第三方估算、模型推断或单一案例包装成事实。
工作流
1. 拆分任务
把评论聚类、参数对比、关键词归类、专利检索等重复劳动与立项、备货和差异化判断分开。
2. 追问口径
对每个结论追溯数据源、时间窗、站点、样本、直接竞品集合和计算方法；无法追溯的结论不进入决策。
3. 校验竞品
用购买对象、场景、关键属性和价格带定义直接竞品，不能把搜索结果总数当作竞争者数量。
4. 寻找异常值
共性痛点用于基础需求，少数但高价值的特殊场景作为待验证假设；AI 不得自动把低频反馈升级为需求事实。
5. 设置人类闸门
在产品立项、供应商下单、预算、广告和发布前输出证据包、反方解释和待批准事项。
6. 回写经验
将真实结果、失败原因和规则修正回到案例库，持续评估 Agent 的命中率与校准度。
判断标准
- 完整、流畅的报告不等于可靠结论；可验证性优先于表达质量。
- AI 可提出推荐，但最终结论必须包含反证、未知项和人工签字点。
- 低频评论只能成为探索线索，需通过关键词、竞品、访谈或测试交叉验证。
- 自动化质量用后验结果评估，不以执行速度或报告篇幅评估。
第三方 MCP 数据
需要外部关键词、竞品、评论或公开网页证据时，读取 references/mcp-data-plan.md，并使用 scripts/mcp_research.py。
- 先动态执行
tools/list、search-tools和describe，依据实时inputSchema构造参数。 - 凭证只从环境变量读取，不进入参数、URL、Skill、终端输出或 Git。
- 可能计费的
tools/call先展示 Provider、工具、无密钥参数、预计成本与输出位置，核对已有授权；仅在授权覆盖本次范围时使用--allow-cost，该标志不是费用上限。 - 脱敏结果用
--output写入 Skill 包之外的任务私有目录；不假设安装位置受仓库.gitignore保护。第三方数据标为估算或代理证据。 - 失败一次后记录缺口，不以重复付费重试掩盖不可用状态。
必须交付的结果
- 自动化/人工责任矩阵
- 证据血缘表
- 结论反证清单
- 人工审批卡
- Agent 后验评估方案
结尾列出站点、数据窗口、证据来源、关键假设、缺口、风险、下一步和所有待批准动作。证据不足时写 HOLD，不得包装成可直接执行。
执行细节、证据字段和质量检查见 references/playbook.md。
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 91 lines · 52 tokens per session scan A d055adec021e
sealeap-xiezhi-amazon-ai-product-research-governance is a skill published in the GitHub repository xjli360/sealeap-amazon-ad-skills (86 stars, last pushed 6d ago), licensed MIT. It adds 52 tokens to every session and 1,222 once invoked, about $0.0003 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 2026-09-07.
Other skills, from other repositories
zach-product-research
A Sorftime-based product research skill for finding Amazon market opportunities and deciding whether a new product is worth pursuing. Sorftime is a market-research data source for Amazon sellers.
zach-listing-health-checker
An Amazon listing health checker that examines a product page as a shopper would see it, including visibility, price, seller, cart, delivery, category, rank, reviews, and search visibility.
zach-sif-cvr-threshold-analyzer
An analyzer that compares daily conversion data for an Amazon product with its natural keyword rankings. CVR, or conversion rate, is the share of visitors who buy; natural rankings are unpaid search positions.
creative-generate
A workflow for creating advertising images and banners from a strategy brief. It separates the main visual artwork from the accurately typeset text and produces variants for display or social formats.
ecom-social
Instructions for creating and improving e-commerce social-media content, advertising, and community work across platforms such as Instagram, YouTube, TikTok, Pinterest, Reddit, WhatsApp, and Xiaohongshu.
ecom-compliance
Check product compliance, HS codes, IP risks, and platform requirements. Use for category approval, FDA/FCC/CE documentation, IP infringement screening, or dangerous goods classification.