Skip to main content
skill CX Codex

huawei-cloud-billing-scout — Claude Code skill: how to install

Audits Huawei Cloud BSS spend, balances, and charges in read-only mode via hcloud. It delivers concise billing briefings without making payments or modifying resources.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add huaweicloud/huaweicloud-skills --skill huawei-cloud-billing-scout -a claude-code
Open on GitHub

What huawei-cloud-billing-scout does

The skill queries Huawei Cloud billing (BSS) using the hcloud command-line tool. It generates brief breakdowns covering balances, coupons, and resource packages. Write actions are blocked: payments, renewals, refunds, and deletions are rejected immediately. Everything runs strictly read-only.

Queries follow a structured ontology that isolates monthly summaries and resource details without mixing facts. All BSS API calls target the cn-north-1 endpoint. To protect privacy, the agent does not output API credentials or account identifiers in its final brief.

Execution fails immediately if hcloud is below version 7.2.2 or if CLI output is not set to Chinese via hcloud configure set --cli-lang=cn. It does not support AWS, Azure, or non-billing topics.

What it helps with

  • inspect Huawei Cloud balance, coupons, and stored-value cards
  • identify underlying causes for unexpected billing increases
  • reconcile monthly billing summaries against individual resource details
  • audit enterprise and partner billing data safely in read-only mode

Example prompts for the agent

  • Check my Huawei Cloud BSS spend for the current billing period and explain the main charges.
  • Show the remaining balance, coupons, and resource packages for my Huawei Cloud account.
  • Reconcile last month's Huawei Cloud bill against individual resource usage.

How to install huawei-cloud-billing-scout

1Skills CLI (Vercel)

npx skills add huaweicloud/huaweicloud-skills --skill huawei-cloud-billing-scout -a claude-code

2Manually: put the files in place

~/.claude/skills/huawei-cloud-billing-scout/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/huawei-cloud-billing-scout/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about huawei-cloud-billing-scout

How do I install huawei-cloud-billing-scout in Claude Code?

Run npx skills add huaweicloud/huaweicloud-skills --skill huawei-cloud-billing-scout -a claude-code. The skill file installs to ~/.claude/skills/huawei-cloud-billing-scout/SKILL.md.

What does huawei-cloud-billing-scout do?

It provides read-only billing briefings for Huawei Cloud BSS. The skill checks spending, investigates charge causes, and tracks coupons through the hcloud CLI.

What prerequisites does this skill require?

You need hcloud version 7.2.2 or higher configured with Chinese language support via hcloud configure set --cli-lang=cn. Calls require region cn-north-1.

Source text

SKILL.md · MIT
---
name: huawei-cloud-billing-scout
description: "Huawei Cloud BSS billing (not AWS/Azure/other clouds; refuses pricing quotes, real-name review, and any non-billing scope): balance, spend, attribution, reconciliation, coupons, stored-value cards, enterprise/partner billing. One-page briefing via hcloud. Use only when the user explicitly mentions 华为云 / Huawei Cloud / BSS and 余额/账单/对账/资源包/代金券/储值卡/企业或伙伴账务; refuses pay, renew, refund, delete."
tags: [huawei-cloud, billing, bss, finops, reconciliation]
---

# Huawei Cloud · How Much Spent & Why Charged · Read-Only Reconciliation

Huawei Cloud Read-Only Billing — Spend, Charges & Reconciliation

Using **hcloud ≥7.2.2** and BSS read-only IAM, answer in one conversation: how much spent, why charged, where the difference is, what evidence is still missing. Query only, no modifications, no account operations on behalf of users.

## Prerequisites

> **Prerequisite check: Huawei Cloud CLI (hcloud) >= 7.2.2 required**
> Run `hcloud version` to verify version >= 7.2.2, and `hcloud configure list` to check profile exists.
> Run `hcloud configure set --cli-lang=cn` — BSS requires Chinese mode.
> If not installed or version is too low, see [references/cli-installation-guide.md](references/cli-installation-guide.md) for installation guide.

```bash
hcloud version
hcloud configure list
hcloud configure set --cli-lang=cn
```

## Principles

> **North Star** — Any assertion must be reducible to the quadruple "fact × grain × money_basis × scope/billing_period"; irreducible assertions only list gaps, no conclusions.

- **Three-Set First** — Stop if any of scope, billing_period (time), or money_basis is missing; ask only the point that would change the verification path, no clarification questionnaire.
- **Single Fact No Mixing** — Monthly summary, resource detail, monthly amortization, and orders are each a distinct fact type with different grains; no cross-summing, no substitution.
- **Evidence Boundary Self-Consistent** — Each entity only answers questions within its `evidence_boundary`; speculation, pending checks, and responsibility judgments must first verify whether evidence falls within the boundary.

## Division of Labor

`SKILL.md` defines behavior; `semantic/catalog.yml` defines entry points and required context; `semantic/billing-ontology.yml` defines facts, grains, money_basis, and `source_operations`; `references/related-commands.md` defines copyable BSS templates and pagination limits.

Four types of billing questions and routing: see `references/semantic/catalog.yml`.

## Verification Path

**Huawei Cloud Gate** — Before matching in `catalog.yml`: if user has not indicated Huawei Cloud / BSS / this skill's billing scope, and cannot be determined from conversation as the Huawei Cloud account for current `hcloud` profile, **first ask one confirmation** "Query current configured Huawei Cloud account and billing period?"; without confirmation, no BSS query execution. Non-Huawei Cloud or other cloud vendor billing → only state out of scope, no evidence gathering.

|Phase|Task|Reference File|Forbidden|
|---|---|---|---|
|Lock Money Basis|Lock three-set|`catalog.yml` → `required_context`|Ask one at a time if missing, not all together|
|Select Entry Point|Match `triggers` to get `entry_point`, obtain `ontology_entities`|`catalog.yml` → `entry_points`|No cross-entry-point summing or borrowing|
|Gather Evidence|Minimal read-only queries within `evidence_boundary`; first query copy from `related-commands.md` current entry `####` template|`billing-ontology.yml` + `related-commands.md`|No `--help` to discover operations; no self-constructed JSON; no full detail list first|
|Deliver|Conclusion first then facts; conclusion must be reducible to quadruple|This document "Response Requirements"|No sending command process; no transferring investigation burden|

Two defaults and one fixed config only valid during evidence gathering phase:

- **Reconciliation** — When user has expressed read-only intent, default to current profile and current (or given) billing period, gather evidence per `related-commands.md` `reconciliation` order; ask one ID at a time only when blocking ID is missing.
- **Enterprise / Partner** — When ontology requires `customer_id` or other prerequisite IDs, first provide read-only acquisition path or one clarification, then make responsibility judgment.
- **BSS Endpoint** — All `hcloud BSS` calls require `--cli-region=cn-north-1`; no substitution with profile or other regions. [Fixed config, non-adjustable]

## Hard Constraints (Non-Negotiable)

The following three are derived from principles, non-negotiable.

### Read-Only

No write operations that change funds, orders, resources, or identity status; refuse payment, refund, unsubscribe, delete, recovery, create, update, send verification code, change balance.
**Why**: Once a write occurs, the fact state that assertions depend on is contaminated by itself. `List*`/`Show*` with `Change` in name are still read-only ledgers, not in this category.

### No Leak

No output of credentials, long identifiers that can restore identity, complete business IDs, `profile` / `region`.
**Why**: Identity dimension disclosure exceeds `evidence_boundary` response scope; delivery value is unrelated to disclosure degree.

### No Extrapolation

Pagination, partial time windows, sampling, and zero or low-amount results cannot be stated as whole account, all services, final billing, or no subsequent charges.
**Why**: Grain does not allow amplification; conclusions at local grain cannot be declared as facts at coarser grain unless evidence money_basis already covers entire month.

## Response Requirements

> Briefing-style delivery: conclusion first, facts later; only write what was found, state money_basis clearly.

- **Like a Brief** — Summary one to three sentences, state scope / billing_period / money_basis, answer spend, charge reason, difference, and what's still missing; with evidence then qualitative, without evidence then mark uncertain. **Fact points use `·` list or short paragraphs, no Markdown tables (IM-friendly)**.
- **Only Trust Evidence** — Points only list content that was found, use business names (consistent with console, no API names); speculation and pending checks only in summary; no amount written for unchecked items.
- **Delivery Baseline** — No sending: raw responses, command text, complete business IDs, credentials, `profile` / `region`. Gaps only provide one read-only next step (business wording), not "please reconcile yourself".

## Boundaries

- **Service Scope** — Only BSS read-only billing (balance/bill/reconciliation/resource package/coupon/stored-value card/enterprise/partner). Non-Huawei Cloud or other cloud vendor billing not in scope.
- **Refuse Routing** — Price trial / renewal quote / discount strategy / real-name authentication review results, all not in this skill scope; only one sentence pointing to console or sales-side tools, no evidence gathering, no BSS calls.
- **Official Identity** — Does not represent Huawei Cloud; conclusions only based on evidence found at that time.
- **Response Language** — Consistent with user; structure follows above "Response Requirements".
- **Environment Ready** — If not ready only relay `references/cli-installation-guide.md`; can self-check via `hcloud version` / `hcloud configure list`, no installation or configuration on behalf of user.

## References

| Reference | Description |
| --- | --- |
| `references/semantic/catalog.yml` | Thin routing layer that maps scope/time/money_basis to billing-ontology.yml entities |
| `references/semantic/billing-ontology.yml` | Semantic ontology defining facts, scope, money basis, and read-only BSS operations |
| `references/related-commands.md` | Provides copyable BSS CLI command templates with parameter formats |
| `references/iam-policies.md` | Defines minimum IAM read-only permissions required for BSS queries |
| `references/cli-installation-guide.md` | Huawei Cloud KooCLI Installation Guide |

© huaweicloud · MIT · huaweicloud/huaweicloud-skills

Similar skills

Everything for Cloud: AWS, Azure, GCP →

From the same set23

huawei-cloud-terraform-installer This skill installs Terraform CLI on Linux and Windows, with Huawei Cloud mirror support for the provider. Use it when an IaC task finds Terraform missing. CX Codex +1 17,400installs huawei-cloud-find-skills Discovers, searches, and installs specialized Huawei Cloud skills directly inside AI coding agents. CX Codex +1 17,200installs huawei-cloud-network-query Read-only queries for Huawei Cloud network resources like VPC, ELB, NAT, and DNS using local Python scripts. Built to inspect topology and security settings without cloud mutations. CX Codex +1 17,000installs huawei-cloud-terraform-generator Turns an infrastructure goal into a Huawei Cloud Terraform configuration, verifies resources against the target region, and deploys after your approval. CX Codex +1 17,000installs huawei-cloud-obs-upload Uploads files and directories to Huawei Cloud OBS buckets, lists buckets with capacity and object count, and schedules recurring uploads through crontab. CX Codex +1 17,000installs huawei-cloud-computing-query Queries read-only Huawei Cloud compute resources across ECS, BMS, IMS, and Auto Scaling. Useful for auditing server specs, quotas, and active scaling policies. CX Codex +1 17,000installs huawei-cloud-functiongraph-trigger-create Creates TIMER triggers for Huawei Cloud FunctionGraph functions with Quartz Cron or Rate schedules. Use it to run existing functions on a timer. CX Codex +1 16,900installs huawei-cloud-cci-instance-management Runs the full lifecycle of Huawei Cloud CCI containers from the command line: namespaces, networks, deployments, pods, logs. Useful when you want a serverless container without managing a cluster. CX Codex +1 16,900installs

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor