Skip to content

huawei-cloud-billing-scout.md

Latest commit

 

History

History
160 lines (122 loc) · 7.23 KB

huawei-cloud-billing-scout.md

File metadata and controls

160 lines (122 loc) · 7.23 KB
Raw

华为云 · 花多少为何扣 · 只读对账

huawei-cloud-billing-scout · Huawei Cloud Read-Only Billing — Spend, Charges & Reconciliation

Read-only Huawei Cloud / 华为云 BSS FinOps via KooCLI (hcloud ≥7.2): balance, monthly spend, charge attribution, reconciliation, coupons, enterprise and partner billing. One-page briefing—how much, why charged, what differs, what evidence is still missing.

华为社区版 · 社区维护，非华为云官方；结论以当次 hcloud/BSS 响应为准。

Version: 2.3.9 · Changelog: qa/huawei-cloud-billing-scout/CHANGELOG.md · Pre-order pricing: huawei-cloud-cost-estimation.md

What it does

Capability Typical questions
Account facts Balance, debt, stored-value cards, monthly spend, ledgers
Charge attribution Top spenders, resource charges, charges after delete, usage and amortized cost
Reconciliation Console vs export, summary vs detail, order vs usage
Entitlements Resource packages, coupons, partner quotas, deduction gaps
Scope Enterprise/sub-account, partner/reseller views
Out of scope Pricing quotes → use huawei-cloud-cost-estimation; real-name review; non-Huawei-Cloud billing

Runtime bundle (install payload)

Only this tree is copied by npx skills add or Hermes local sync:

skills/huawei-cloud-billing-scout/
├── SKILL.md                 # 原则 · 分工 · 查证路径（四阶段表）· 红线 · 答复 · 边界
└── references/
    ├── related-commands.md
    ├── cli-installation.md
    ├── iam-policies.md
    └── semantic/
        ├── catalog.yml
        └── billing-ontology.yml

No evals/, qa/, gate.py, lint configs, or *-workspace/ under skills/.

In-skill flow

User question
     │
     ▼
SKILL.md 华为云门禁 ─── confirm Huawei Cloud scope / profile when unclear
     │
     ▼
catalog.yml ─── triggers (华为云-prefixed) → ontology_entities
     │
     ▼
billing-ontology.yml ─── fact, scope, money_basis, evidence_boundary
     │
     ▼
related-commands.md ─── smallest read-only hcloud BSS query set
     │
     ▼
答复格式 ─── 小结 → 事实要点 → 一条只读下一步（如有）

53 unique read-only BSS query operations (KooCLI 7.2.2), aligned across ontology, commands, and qa/.../fixtures/ops_contracts.yml. Pricing-quote and real-name review operations were removed in 2.3.7 to retract the skill to a billing-only boundary; see out_of_scope in catalog.yml.

SKILL.md structure

Section Purpose
原则 北极星 + 三件套（scope/账期/口径）+ 单一事实不混 + 证据边界自洽
分工 SKILL.md / catalog.yml / billing-ontology.yml / related-commands.md 四件套各司其职
查证路径 华为云门禁 + 四阶段表（定口径 → 选入口 → 取证 → 交付），含对账与企业/伙伴默认
红线 只读 / 不泄密 / 不外推，三条各附一句「为何」从原则派生
答复 答复格式 briefing delivery contract（见下）
边界 服务范围（仅 BSS 只读账务）、拒绝路由（报价/实名/非华为云）、华为社区版声明、答复语言、环境就绪

答复格式 (briefing-style output contract)

Answers follow 答复格式 in SKILL.md (conclusion-first summary, then fact bullets):

  1. Briefing summary (1–3 sentences) — scope, billing period, money basis; spend / cause / delta / gaps.
  2. Fact bullets — only queried facts; 业务称呼 (console-aligned; no API names in user text).
  3. Delivery floor — no JSON walls, command logs, full business IDs, credentials, or profile/region; at most one read-only next step in plain language; never「请自行对账」.

IM-safe: facts are rendered as · lists or short paragraphs, not Markdown tables — the delivery channel is often IM (Feishu, WeChat, etc.) where pipe characters do not render.

Grading: qa/huawei-cloud-billing-scout/evals/llm-rubric.yml (merged into all 24 eval cases; evals 22–24 exercise dialogue-progression discipline — "一次一问").

Safety boundary

Allowed: BSS List* / Show* queries; product List/Show/Get only after BSS identifies a resource.
Refused: pay, renew, refund, unsubscribe, create, update, delete, reclaim, transfer, or any mutation.

ListCustomer*ChangeRecords are read ledgers (Change ≠ write). No auto-install of hcloud. Real BSS only when maintainers set HUAWEICLOUD_BILLING_SCOUT_REAL=1.

QA (not installed with skill)

qa/huawei-cloud-billing-scout/
├── validate.sh
├── skillcheck.toml
├── .markdownlint.json
├── policy.skill-scanner.yaml
├── evals/evals.json
├── evals/llm-rubric.yml
├── fixtures/ops_contracts.yml
└── bin/
    ├── gate.py
    ├── verify_ops.py
    └── export_llm_eval.py

./qa/huawei-cloud-billing-scout/validate.sh
python3 qa/huawei-cloud-billing-scout/bin/gate.py style   # optional style-only

Default gate is offline. It does not call BSS unless HUAWEICLOUD_BILLING_SCOUT_REAL=1 is set for maintainer smoke tests.

Install

skills.sh

npx skills add ontology-of-everything/SemanticSkills \
  --skill huawei-cloud-billing-scout \
  --agent cursor \
  --copy -y

Agents: cursor, claude-code, codex — see docs/agents/. Hermes: hermes.md.

Local development:

npx skills add ./skills/huawei-cloud-billing-scout \
  --skill huawei-cloud-billing-scout \
  --agent cursor \
  --copy -y

Marketplaces

  • skills.sh
  • SkillsMP — repo topics claude-skills, claude-code-skill; verify with curl "https://skillsmp.com/api/v1/skills/search?q=huawei-cloud-billing-scout"
  • ClawHub — publish from skills/huawei-cloud-billing-scout/ after ./qa/.../validate.sh

ClawHub publish (only after explicit release approval):

# Use absolute path to the skill folder (relative ./skills/... may fail on some CLI versions)
clawhub skill publish "$PWD/skills/huawei-cloud-billing-scout" \
  --slug huawei-cloud-billing-scout \
  --name "Huawei Cloud Read-Only Billing — Spend, Charges & Reconciliation" \
  --version <semver> \
  --changelog "<semver>: ClawHub audit alignment — drop pricing-quote and real-name review (retract to billing-only boundary), Huawei Cloud-prefixed catalog triggers, provider_gate" \
  --clawscan-note "Huawei Cloud BSS List/Show read-only only. Billing scope only (balance/spend/attribution/reconciliation/coupons/cards/enterprise/partner); refuses pricing-quote, real-name review, and non-Huawei-Cloud billing; agent must not install hcloud; no writes" \
  --tags latest

ClawHub skill bundle is MIT-0; repository source remains Apache-2.0. Installable SKILL.md must not declare a conflicting license in frontmatter.