菜单
Investigates individual users and user populations in Amplitude — resolve users by email/ID, read profiles, batch-analyze event timelines, spot-check cohort membership. Use for "what did user X do", attribution audits, population sampling, and email-to-user-ID resolution.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/amplitude/mcp-marketplace --skill user-cohort-forensics命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Summarizes B2B account health by analyzing usage patterns, engagement trends, risk signals, and expansion opportunities. Use for customer success reviews, renewal preparation, QBRs, or account prioritization.
Analyzes what users ask AI agents about and how well each topic is served. Only use when the user has Amplitude Agent Analytics instrumented in their project. Use when the user asks "what are people asking the AI", "top AI topics", "where is the AI struggling", "AI coverage gaps", "what should we improve in our AI", or wants product insights from AI conversation patterns.
Deeply analyze Amplitude dashboards by analyzing key charts, surfacing top areas for concern and takeaways, identify anomalies, then explain changes using customer feedback trends
正在显示 SKILL.md
| name | user-cohort-forensics |
| description | Investigates individual users and user populations in Amplitude — resolve users by email/ID, read profiles, batch-analyze event timelines, spot-check cohort membership. Use for "what did user X do", attribution audits, population sampling, and email-to-user-ID resolution. |
Single user deep-dive: get_amp_user_data include: 'id' (resolve the
user: email, user ID, device ID, or amplitude ID) → include: 'profile'
(lifecycle, acquisition, usage stats) → include: 'timeline' (session-aware
event history). include: 'both' gets profile + timeline in one call when
you know you'll need both. Summarize journeys; don't dump raw events.
Population analysis (batched): build the user set first
(query_amplitude_data with a user-ID group_by to rank by volume, or
use_amplitude_cohorts action: 'find' for a filtered set) →
get_amp_user_data include: 'timeline' in parallel batches — up to 10
identifiers per call, 10–20 calls in flight is normal for population
analysis; hundreds of calls total is fine. Keep each call narrow (event
types, window) so responses stay small.
Cohort spot-check: prefer existing cohorts — use_amplitude_cohorts
action: 'list' or action: 'get' before building ad
hoc definitions. action: 'membership' verifies specific users. Check a
member's timeline for the exact markers (purchase, typing, checkout events)
rather than trusting the cohort definition blindly.
Email → ID resolution: get_amp_user_data include: 'id' per email;
uploaded email lists are resolved in batches of ≤10 identifiers per call.
For bulk exports, batch and note the retry pattern on individual failures.
include: 'timeline': always bound the window (last 30 days by default)
and pass event-type filters when you know what you're looking for —
unfiltered timelines are large and slow.retryable with retryAfterMs —
back off instead of churning.