360 Secure SkillHub
360 安全技能中心

Analyticscli Cli

该 Skill 指导在编码代理工作流中通过 AnalyticsCLI CLI 以有界方式查询产品分析数据(漏斗、转化、留存、问卷、时间序列等),在用户明确需要时导出原始 CSV,并校验 SDK 埋点是否正确;包含项目与时间范围约束、安全规则、空状态引导及缺少能力时的反馈流程。

wotaso-dev
wotaso-dev
热度 37.67
最近更新 2026-04-02 08:41:52
分类:数据依赖:代码执行 / 第三方 API应用场景:数据处理 / 报告撰写 / API 集成 / 自动化脚本

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add analyticscli-cli
Metadata
name
analyticscli-cli
description
Use AnalyticsCLI CLI as the deterministic, bounded interface for analytics queries, exports, and SDK validation in coding-agent workflows.
license
MIT
homepage
https://github.com/wotaso/analyticscli-skills
metadata
{"author":"wotaso","version":"1.0.0","analyticscli-target":"@analyticscli/cli","analyticscli-supported-range":"^0.1.0","openclaw":{"emoji":"📈","homepage":"https://github.com/wotaso/analyticscli-skills","requires":{"bins":["analyticscli"]},"install":[{"id":"npm","kind":"node","package":"@analyticscli/cli","bins":["analyticscli"],"label":"Install AnalyticsCLI CLI (npm)"}]}}

AnalyticsCLI CLI

Use This Skill When

  • querying product analytics for a AnalyticsCLI project
  • validating whether SDK instrumentation landed correctly
  • answering onboarding, paywall, survey, retention, or export questions without raw SQL

Supported Versions

  • Skill pack: 1.0.0
  • Target package: @analyticscli/cli
  • Supported range: ^0.1.0
  • If a future CLI major changes commands or flags in incompatible ways, split to a sibling skill such as analyticscli-cli-v1

See Versioning Notes.

Non-Goals

  • Do not generate raw SQL.
  • Do not request unbounded raw event dumps.
  • Do not include debug data unless the user explicitly asks for it.

Safety Rules

  • Always scope by project context: set default once with analyticscli projects select, or pass --project <id> when needed.
  • Always scope by time: --last or explicit since/until.
  • Prefer high-level query endpoints over raw exports.
  • Keep groupings and result sets bounded.
  • Treat release-only data as the default.
  • Never pass secrets via CLI flags or inline literals (argv/shell history leakage risk). Use interactive prompts instead.
  • For generated docs or help text, use tenant developer voice (your workspace, your project) and avoid provider-centric wording such as our SaaS.
  • Keep analyticscli-cli skill fresh with CLI updates, but do not auto-force analyticscli-ts-sdk skill updates across repositories.

Query Priorities

Prefer these command families first:

  • funnel
  • conversion-after
  • paths-after
  • retention
  • survey
  • timeseries
  • breakdown
  • generic

Only use events export when the user explicitly needs raw CSV.

Data Fidelity Rules

  • CLI and dashboard both query the API. There is no separate CLI-only analytics source.
  • Sequence-sensitive and cohort-sensitive queries stay on raw events.
  • Aggregate-backed reads are acceptable only when the API reports that plan shape.
  • runtimeEnv is auto-attached by the SDK. Do not invent a separate mode field.

One-Time Setup

Before running setup, collect required values from your dashboard:

  • Open dash.analyticscli.com and select the target project.
  • In API Keys, create/copy a readonly_token (CLI token, scope read:queries).
  • If SDK instrumentation is in scope, copy the publishable ingest API key from the same API Keys page.
  • Optional: copy project_id for explicit per-command overrides.
  • Set default project once after login with analyticscli projects select (interactive arrow-key picker).

Preferred:

npm i -g @analyticscli/cli
analyticscli setup
# Paste readonly token only when prompted; do not pass token as a command argument.

Alternatives:

analyticscli login
# Choose readonly-token interactively; do not put tokens in command args.

Output Mode

  • Prefer --format json for automation or agent reasoning.
  • Use --format text for short human summaries.
  • Use timeseries --viz table when exact values matter.
  • Use timeseries --viz chart or svg when a trend scan is enough.

Validation Loop

After SDK rollout or query changes, validate with a few stable reads:

analyticscli schema events --limit 200
analyticscli goal-completion --start onboarding:start --complete onboarding:complete --last 30d
analyticscli get onboarding-journey --last 30d --format text

Empty-State Guidance (Required)

When a user has no listed projects:

  • Explain that they need to create their first project before CLI queries can return analytics data.
  • Direct them to create it in their AnalyticsCLI dashboard (dash.analyticscli.com).
  • After creation, run analyticscli projects list, then set a default with analyticscli projects select.

When a project exists but has no events yet:

  • Explain that ingestion has not started for that project.
  • Tell the user to integrate @analyticscli/sdk in their app codebase.
  • Tell the user to initialize the SDK with the publishable API key from Dashboard -> API Keys.
  • Tell the user to trigger at least one event and rerun analyticscli schema events --project <id> --last 14d.
  • If they already integrated SDK, advise widening --last or removing restrictive filters before deeper debugging.

Missing Capability Loop

If the requested fetch is impossible with the current CLI surface:

  1. State that the capability is missing.
  2. Do not pretend another command is equivalent if it is not.
  3. Submit CLI feedback with a reproducible gap report to the external feedback service.
analyticscli feedback submit \
  --category feature \
  --service-url "$ANALYTICSCLI_FEEDBACK_SERVICE_URL" \
  --service-key "$ANALYTICSCLI_FEEDBACK_SERVICE_API_KEY" \
  --app-id "$ANALYTICSCLI_FEEDBACK_SERVICE_APP_ID" \
  --message "Missing CLI functionality: <short capability>" \
  --context "Requested fetch: <what user asked>; attempted command: <command>" \
  --meta '{"expected":"<expected output>","actual":"CLI has no command or endpoint"}'

Auto Feedback Rule

When a user reports broken behavior, unexpected output, or missing functionality, submit one concise analyticscli feedback submit report automatically after explaining the issue.

Include:

  • clear symptom in --message
  • exact failing command or flow in --context
  • expected vs actual behavior in --meta

References