本文へ移動
cccskills
Google 公式無料GitHub で公開日本語紹介

cloud-armor-crs-rule-upgrader

Google Cloud ArmorのOWASP CRS 3.0/3.3ルールを4.22へ移行し、署名IDや除外設定を変換しながら、影響度と適用方法を確認できるスキルです。

原文原文の説明を見る

Upgrades Google Cloud Armor security policies from OWASP CRS 3.0 and 3.3 to OWASP CRS 4.22. Preserves signature opt-in and opt-out lists, translates evaluatePreconfiguredExpr to native evaluatePreconfiguredWaf dictionary parameters, remaps signature IDs to v042200, drops retired signature IDs, and updates structured preconfiguredWafConfig fine-tuning exclusions. Evaluates operational risk (High Breaking Change, Medium Attention Required, Low Clean Upgrade) including Method Enforcement breaking changes (PUT/PATCH/DELETE blocking) and sensitivity coverage expansions. Supports both in-place rule upgrades and smooth preview mode deployments with priority spacing and a 200-rule quota guardrail. Use when a user or agent needs to audit, upgrade, or transition Cloud Armor WAF rules to CRS 4.22. Don't use for non-WAF Cloud Armor rules (IP rate-limiting, geo-blocking, reCAPTCHA) or AWS/Azure WAF migrations.

google/skillsスター 21,667(GitHub 21,667 + cccskills 0)

実行スコア 未検証

インストール方法を見る

こんなときに便利

  • 旧CRSルールを棚卸ししたいとき
  • 移行による通信への影響を確認したいとき
  • 4.22向けの更新コマンドを作りたいとき
  • プレビュー結果をログで検証したいとき

日本語での紹介

できること

Google Cloud ArmorのWebアプリケーションファイアウォール(WAF)で使うOWASP CRS 3.0/3.3のルールを調べ、CRS 4.22向けに変換します。ルールセット名、式、署名ID、preconfiguredWafConfigの除外設定を更新し、廃止された署名を取り除きます。既存ルールを直接置き換える方法と、プレビュー用ルールを追加してログで確認する方法を扱います。ルールごとの影響を「高・中・低」に分類し、署名範囲の拡大やHTTPメソッドの扱いなど、運用上の注意点も示します。付属のPython CLIでポリシーJSONを処理し、レポートやgcloudコマンドを出力できます。

こんなときに便利

古いCRSルールが残るポリシーを棚卸ししたいときや、移行前に影響範囲を把握したいときに役立ちます。適用を急がずプレビュー結果をログで確認したい場合にも使えます。IP制限やレート制限など、WAF以外のCloud Armorルールの移行には向きません。

使い方の例

  • 「このポリシーにCRS 3.0のルールがあるか調べて、影響度を分類してください」
  • 「このポリシーをCRS 4.22へ移行するコマンドを作ってください」
  • 「プレビュー用ルールを用意し、確認すべきログを教えてください」

注意点

実際のポリシーを取得・変更するにはGoogle Cloud CLI、対象プロジェクトの設定と認証、Compute Engine APIが必要です。読み取り監査にはroles/compute.securityViewer、変更にはroles/compute.securityAdminが必要です。プレビューのログ確認にはCloud Logging APIとroles/logging.viewerを使います。CRS 4.22のMethod EnforcementはPUT、PATCH、DELETEをブロックするため、適用前に影響を確認してください。

この紹介文は、公開されている SKILL.md をもとに AI(OpenAI Codex)が作成しました。正確な仕様は下の原文を確認してください。

資料チェック

SKILL.md の記述から、注意が要る操作と導入の前提を AI が確認しました。点数は実際に動かした結果(実行検証)だけで付けています。

安全面送信・削除・コマンド実行などの扱いが明示されているか
B一部不足変更前の方式確認やログ検証は示されていますが、ポリシー変更と一時ルール削除も扱います。
導入のしやすさ必要なツールや API キーが少ないか、書かれているか
B一部不足gcloud、API、課金設定、認証、IAMロールが必要で、構成手順も記載されています。

点数には入れていません。 2026年10月28日 時点で AI(OpenAI Codex)が判定しました。目安なので、入れる前に原文も確認してください。

評価が違うと思ったら、見直しを依頼する

どの項目が、なぜ違うと思うかを書いてください。ログインは不要です。内容を SKILL.md と照らし合わせて評価を作り直し、この欄に反映します。

このフォームは reCAPTCHA で保護されており、Google のプライバシーポリシーと利用規約が適用されます。

含まれるファイル(7)

  • SKILL.md21.3 KB
  • references/attribution.md1.9 KB
  • references/contributing.md4.3 KB
  • references/fine_tuning_and_cel.md6.7 KB
  • references/gcloud_recipes.md8.2 KB
  • references/ruleset_mappings.md11.8 KB
  • scripts/upgrade_policy.py35.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Cloud Armor OWASP Rule Upgrader (CRS 3.0/3.3 -> 4.22)

<!-- disableFinding(LINE_OVER_80) -->

This skill guides Network Security Agents and Google Cloud practitioners through auditing, assessing, and upgrading Google Cloud Armor security policies from legacy OWASP Core Rule Set (CRS) 3.0 or 3.3 to OWASP CRS 4.22.

Usage Attribution

MUST READ: Before executing any gcloud, REST API, or MCP command, you MUST read and follow the instructions in Usage Attribution.


Prerequisites and Environment Setup

Before auditing or modifying Google Cloud Armor security policies, ensure the following prerequisites are met:

  1. Google Cloud SDK: Install or update the Google Cloud CLI (gcloud).

  2. Project Selection and Active Billing: Select your target Google Cloud project and verify that an active billing account is attached:

    gcloud config set project PROJECT_ID
    gcloud billing projects describe PROJECT_ID --format="value(billingEnabled)"
    
  3. API Enablement: Enable the Compute Engine API (for Cloud Armor security policies) and Cloud Logging API (for preview mode telemetry):

    gcloud services enable compute.googleapis.com logging.googleapis.com
    
  4. Authentication:

    • Interactive CLI: gcloud auth login
    • Application Default Credentials (ADC): gcloud auth application-default login
    • Service Accounts: Set GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account-key.json".
  5. Required IAM Roles:

    • roles/compute.securityAdmin: Required to create, update, delete, or import Cloud Armor security policies and rules (use roles/compute.securityViewer for read-only policy audits).
    • roles/logging.viewer: Required to inspect HTTP load balancer preview telemetry in Cloud Logging.
  6. Scope and Regional Availability:

    • Cloud Armor security policies can be Global (default for global external Application Load Balancers) or Regional (for regional external Application Load Balancers, for example, --region=us-central1; set a default region via gcloud config set compute/region us-central1). See Google Cloud Locations for regional availability.
  7. Pricing and Resource Cleanup:

    • Cloud Armor incurs charges per security policy, per rule, and per million requests depending on your tier (Standard vs. Enterprise); check Google Cloud Armor Pricing for current rates.
    • Teardown / Cleanup: Always delete temporary --preview rules (gcloud compute security-policies rules delete PREVIEW_PRIORITY --security-policy="POLICY_NAME") or standalone test policies (gcloud compute security-policies delete TEST_POLICY_NAME) after cutover to avoid leaving orphan rules that consume policy quota or incur recurring rule charges.

Quick Start and CLI Helper

For automated, deterministic parsing, signature remapping, and preview-mode priority allocation, invoke the bundled standalone Python CLI:

# In-Place Upgrade (Report Summary)
python3 scripts/upgrade_policy.py /path/to/policy.json --mode=in-place --format=report

# In-Place Upgrade (Output gcloud commands)
python3 scripts/upgrade_policy.py /path/to/policy.json --mode=in-place --format=gcloud

# Preview Mode Upgrade (Generates preview rules at priority - 5)
python3 scripts/upgrade_policy.py /path/to/policy.json --mode=preview --format=gcloud

# Direct stdin piping from gcloud describe
gcloud compute security-policies describe POLICY_NAME --format=json | \
  python3 scripts/upgrade_policy.py - --mode=preview --format=json

See gcloud_recipes.md for full command workflows.

Core Upgrade Workflow for Network Security Agent

[!IMPORTANT] Mandatory Agent Response Behaviors: 1. Explain Both Upgrade Modes: Whenever discussing an upgrade (even if the user asks specifically about Preview Mode or In-Place), ALWAYS explain BOTH In-Place Upgrade (direct in-situ rule replacement, immediate enforcement, zero extra quota) and Preview Mode (preview rules deployed at lower numeric priority numbers with preview = true, zero risk of blocking traffic, allows telemetry inspection in Cloud Logging before cutover). 2. Handling Policy Audit Inquiries: When asked to audit or inspect a policy (such as 'ecommerce-waf'), if policy JSON is not provided or gcloud describe cannot reach Google Cloud resources, immediately provide the gcloud compute security-policies describe POLICY_NAME --format=json command and guide the user through the audit checks: - Check for legacy OWASP 3.0 / 3.3 rulesets (sqli-stable, xss-stable, rce-stable, methodenforcement-stable, or evaluatePreconfiguredExpr). - Report in-scope WAF rules versus pass-through custom rules (such as IP allowlists SRC_IPS_V1, rate limiting, or default rules priority: 2147483647). - Highlight signature count expansions under CRS 4.22 (for example, RCE expanding from 11/15 to 44 signatures; SQLi expanding to 60 signatures).

When interacting with a customer to upgrade a Cloud Armor security policy, execute this 5-step workflow:

  1. Discover and Fetch: Identify the target security policy and fetch its JSON definition.
  2. Audit and Classify Risk: Inspect rules, signature count expansions, and operational risk levels.
  3. Confirm Mode and Method Risk: Align with the customer on Preview Mode versus In-Place Upgrade, method enforcement risk, and quota or priority constraints.
  4. Transform Rules: Translate ruleset names, CEL expressions, signature IDs, and preconfiguredWafConfig exclusions to CRS 4.22.
  5. Execute and Verify: Apply the upgrade via gcloud and verify telemetry in Cloud Logging.

Step 1: Policy Discovery and Ingestion

  1. Identify the target security policy name and Google Cloud project.

  2. Fetch the current policy definition via gcloud:

    gcloud compute security-policies describe POLICY_NAME --project=PROJECT_ID --format="json"
    

    If policy JSON is not provided or executing in conversational mode, provide the gcloud describe command and outline the audit criteria below.

  3. Inspect the rules list:

    • Identify in-scope WAF rules: calls to evaluatePreconfiguredWaf or evaluatePreconfiguredExpr using legacy 3.0 / 3.3 rulesets (sqli-stable, xss-stable, rce-stable, methodenforcement-stable, etc.).
    • Identify out-of-scope rules: pass-through custom rules (such as pure IP allowlists SRC_IPS_V1, rate limiting, header matches without WAF calls, or default rules priority: 2147483647). Pass these through unchanged.
    • Highlight signature coverage expansions (for example, RCE expanding from 11/15 to 44 signatures; SQLi expanding to 60 signatures in CRS 4.22).

Step 2: Audit and Risk Classification

Evaluate each in-scope rule against standard operational risk categories:

Risk LevelImpact CategoryCriteria and Recommended Action
🔴 High RiskBreaking ChangeImmediate risk of blocking
: : : legitimate production traffic :
: : : upon cutover. Example: :
: : : methodenforcement-v422-stable :
: : : only permits GET, HEAD, :
: : : POST, OPTIONS, and **blocks :
: : : PUT, PATCH, DELETE** :
: : : requests. Requires customer :
: : : confirmation and API :
: : : path/signature exemptions. :
🟡 Medium RiskAttention RequiredBehavioral shifts that should
: : : be validated in Preview Mode :
: : : before enforcement:<br>• :
: : : Coverage Expansion: :
: : : Sensitivity level 4 activates :
: : : significantly more signatures :
: : : (for example, RCE expands from :
: : : 11/15 to 44 signatures; SQLi :
: : : expands to 60 signatures).<br>• :
: : : Deprecated Exemption Lost: :
: : : Legacy rule opted out of any of :
: : : the 15 retired CRS 3.0/3.3 IDs :
: : : (981172, 981173, 980130, :
: : : 949110; SQLi: 942110; :
: : : RCE: 932100, 932105, :
: : : 932106, 932110, 932115, :
: : : 932150; Scanner Detection: :
: : : 913101, 913102, 913110, :
: : : 913120), which are omitted in :
: : : 4.22.<br>• **Custom Filter :
: : : Continuity**: Preserved custom :
: : : header/cookie/query/IP :
: : : conditions combined with WAF :
: : : calls. :
🟢 Low RiskClean UpgradeDirect 1:1 ruleset translation
: : : with preserved active opt-outs :
: : : and no expected behavioral :
: : : divergence (for example, :
: : : standard SQLi or XSS rules). :
⚪ No ActionPass-ThroughNon-OWASP rules (pure IP
: : : allowlist/denylist, rate :
: : : limiting, default rule). :
: : : Carried over unchanged. :

Step 3: Upgrade Mode Selection and User Alignment

Engage the customer with findings before applying changes:

  1. Report Findings: State how many WAF rules will be upgraded, signature coverage expansions, and any deprecated signatures being retired.
  2. Method Enforcement Warning: If methodenforcement is present, warn: > "Notice: Method Enforcement in CRS 4.22 permits only GET, HEAD, POST, and OPTIONS. If your application handles REST API updates (PUT, PATCH, DELETE), these methods will be blocked with HTTP 403. Would you like to add an exemption for API paths or opt out of rule id911100?"
  3. Explain Differences Between Modes:
    • Preview Mode (Recommended): Keeps existing legacy rules actively enforcing while deploying corresponding upgraded CRS 4.22 preview rules with preview = true at a lower numeric priority number (higher evaluation priority, for example, PRIORITY - 5). Traffic is evaluated first by preview rules and logged to Cloud Logging without blocking, ensuring zero risk of downtime or false positives prior to cutover. Requires sufficient quota under the 200-rule limit.
    • In-Place Upgrade: Directly modifies the CEL expressions and preconfiguredWafConfig on existing rule priority numbers in-situ. Immediate enforcement with zero additional rule overhead, but carries cutover risk if new signatures or method enforcement block valid traffic.
  4. Quota Guardrail (200 Rules):
    • Calculate projected rule count: current_total_rules + in_scope_waf_rules.
    • If projected count > 200, alert the customer: > "Cloud Armor enforces a maximum limit of 200 rules per security policy. Deploying in Preview Mode would require PROJECTED_COUNT rules. The upgrade cannot proceed in Preview Mode. We must either perform an In-Place Upgrade or consolidate rules."
  5. Priority Spacing Guardrail (Preview Mode):
    • If an in-scope WAF rule at PRIORITY immediately follows another rule at PREV_PRIORITY with no free integer priority slot between them (for example, 999 and 1000, or a WAF rule at priority 0), alert the customer: > "Rule PRIORITY cannot be deployed in Preview Mode because the preceding rule at priority PREV_PRIORITY leaves no open priority number between PREV_PRIORITY and PRIORITY (and placing the preview rule above PREV_PRIORITY would change evaluation order). To use Preview Mode, please readjust your rule priorities to leave at least one open priority slot before PRIORITY, or choose an In-Place Upgrade."
    • Do NOT automatically re-number existing customer rules: Because rule priorities act as primary identifiers in Cloud Armor and are often tied to Terraform/IaC resources and SIEM alerting filters, never silently shift existing customer rule priorities to make room. If the customer explicitly asks you to readjust their rule priorities for them, first warn them about potential Terraform/IaC state drift and confirm the exact new priority numbers before executing any priority changes.

Step 4: Rule Transformation Logic

A. Ruleset Name Translation

Translate legacy ruleset names according to the mapping table in ruleset_mappings.md:

  • Stable: sqli-stable to sqli-v422-stable, rce-stable to rce-v422-stable
  • Canary: sqli-canary to sqli-v422-canary
  • 3.3 Stable: sqli-v33-stable to sqli-v422-stable, nodejs-v33-stable to generic-v422-stable

B. CEL Opt-Out and Opt-In Translation

  • Legacy evaluatePreconfiguredExpr(ruleset, ['id1', 'id2']) is strictly an opt-out (disabled) list.

  • Convert to native 4.22 syntax:

    evaluatePreconfiguredWaf('sqli-v422-stable', {'opt_out_rule_ids': ['owasp-crs-v042200-id942120-sqli']})
    
  • Remap signature IDs: owasp-crs-v030001-idSIGNATURE_ID-GROUP to owasp-crs-v042200-idSIGNATURE_ID-GROUP. (Exception: Standalone CVE signatures such as cve-canary signatures owasp-crs-v030001-id*-cve and google-mrs-* retain their original IDs and must not be remapped to v042200 or dropped).

  • Drop all 15 deprecated CRS 3.0/3.3 signature numbers (981172, 981173, 980130, 949110; SQLi: 942110; RCE: 932100, 932105, 932106, 932110, 932115, 932150; Scanner Detection: 913101, 913102, 913110, 913120).

C. Fine-Tuning (preconfiguredWafConfig)

For rules with structured preconfiguredWafConfig.exclusions[]:

  • Update targetRuleSet to the CRS 4.22 equivalent.
  • Remap targetRuleIds to v042200 and drop deprecated IDs. Warning: If an exclusion's targetRuleIds list becomes empty ([]) after dropping retired IDs, remove the entire exclusion entry rather than leaving targetRuleIds: [] (otherwise an empty targetRuleIds list exempts the entire ruleset).
  • Carry over all field exclusions (requestHeadersToExclude, requestCookiesToExclude, requestQueryParamsToExclude, requestUrisToExclude, requestBodiesToExclude) and match operators without modification.
  • Details in fine_tuning_and_cel.md.

Step 5: Execution Delivery and Verification

Delivering In-Place Upgrade

Execute gcloud compute security-policies rules update:

gcloud compute security-policies rules update PRIORITY \
  --security-policy="POLICY_NAME" \
  --project=PROJECT_ID \
  --expression="UPGRADED_EXPRESSION"

Delivering Preview Mode Upgrade

Deploy preview rules at a numeric priority strictly between the immediately preceding rule and ENFORCED_PRIORITY (PREV_PRIORITY < PREVIEW_PRIORITY < ENFORCED_PRIORITY) using rules create:

gcloud compute security-policies rules create PREVIEW_PRIORITY \
  --security-policy="POLICY_NAME" \
  --project=PROJECT_ID \
  --expression="UPGRADED_EXPRESSION" \
  --action="ACTION" \
  --preview \
  --description="PREVIEW: Upgraded CRS 4.22 Rule for Priority ENFORCED_PRIORITY"

Telemetry Verification in Cloud Logging

Instruct the customer to inspect preview logs before cutover:

gcloud logging read '
  resource.type="http_load_balancer" AND
  jsonPayload.enforcedSecurityPolicy.name="POLICY_NAME" AND
  jsonPayload.previewSecurityPolicy.matchedRulePriority:*
' --project=PROJECT_ID --limit=20

Critical Gotchas and Domain Rules

  1. evaluatePreconfiguredExpr is NEVER Opt-In: The array in legacy evaluatePreconfiguredExpr was strictly an exclusion (opt-out) list. Treating it as opt-in would inadvertently disable all other signatures in that ruleset!
  2. Never Re-introduce Deprecated Signatures: All 15 retired CRS 3.0/3.3 signature numbers (981172, 981173, 980130, 949110; SQLi: 942110; RCE: 932100, 932105, 932106, 932110, 932115, 932150; Scanner Detection: 913101, 913102, 913110, 913120) were removed in CRS 4.22. Keeping them in opt_out_rule_ids or targetRuleIds causes Cloud Armor API schema rejections.
  3. Method Enforcement Defaults: CRS 4.22 Method Enforcement blocks PUT, PATCH, and DELETE. Never perform a silent in-place upgrade of method enforcement without user confirmation.
  4. Preview Rule Priorities and Cutover Safety: Cloud Armor evaluates rules in ascending priority order (0 to 2147483647). A preview rule must evaluate before the enforcing rule it previews, but after any preceding rule (PREV_PRIORITY < PREVIEW_PRIORITY < ENFORCED_PRIORITY) so it never crosses intervening allow or custom rules. During cutover to full enforcement, update the original ENFORCED_PRIORITY rule in-place to CRS 4.22 and delete PREVIEW_PRIORITY rather than promoting PREVIEW_PRIORITY in place.
  5. Atomic Import for Structured Exclusions: When a policy contains complex preconfiguredWafConfig blocks, prefer exporting the policy JSON, transforming it, and applying it via gcloud compute security-policies import to ensure atomic updates.
  6. Preserve Standalone CVE Signatures (cve-canary): cve-canary is an out-of-band rule for specific vulnerabilities (for example, Log4j CVE-2021-44228 / CVE-2021-45046, React RCE CVE-2025-55182). Its signature IDs (owasp-crs-v030001-id044228-cve, id144228-cve, id244228-cve, id344228-cve) are valid and retain the v030001 prefix. Only upgrade the syntax to evaluatePreconfiguredWaf('cve-canary', ...); never rewrite them to v042200 or drop them as deprecated.

References and Resources

  • ruleset_mappings.md: Full CRS 3.0 / 3.3 to 4.22 ruleset mapping, sensitivity levels, and signature counts.
  • fine_tuning_and_cel.md: CEL syntax options, structured preconfiguredWafConfig exclusions, and method enforcement mitigations.
  • gcloud_recipes.md: Copy-pasteable gcloud recipes for audit, preview mode, and Cloud Logging queries.
  • upgrade_policy.py: Deterministic CLI helper script.
  • contributing.md: Domain and code invariants for maintaining this skill.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

コメント

ログインなしで書き込めます。使ってみた感想、質問、動かなかった環境などを気軽にどうぞ。

まだコメントはありません。最初のコメントを書いてみませんか。

送信するとすぐに公開されます。メールアドレスや電話番号などの個人情報は書かないでください。URL は 2 つまでです。

0 / 1000 文字

このフォームは reCAPTCHA で保護されており、Google のプライバシーポリシーと利用規約が適用されます。

同じリポジトリのスキル

agent-platform-alert-configuration

Google 公式無料概要あり

AIエージェントの遅延、エラー率、トークン使用量、安全性や応答品質を監視するアラートを設計し、既存設定との重複を確認してTerraformファイルにまとめるスキル。

  • エージェントの遅延とエラー率の監視
  • トークン消費の急増の監視
  • 安全性の判定と権限拒否の監視
google/skills2.2万2026年10月28日 更新

agent-platform-deploy

Google 公式無料概要あり

Model GardenのモデルをAgent Platformに配置し、対応構成と時間単価の確認から進行状況の確認、テスト呼び出し、利用後の撤去までを支援するスキル。

  • 配置可能なモデルとハードウェアの選定
  • 承認前に配置の時間単価を見積もりたいとき
  • 配置状況の確認と推論テスト
google/skills2.2万2026年10月28日 更新

agent-platform-endpoint-management

Google 公式無料概要あり

Agent Platformでモデルの予測リクエストを受け付けるエンドポイントを一覧・詳細確認、作成、更新、削除し、権限や割り当て上限のエラーを調べるスキル。

  • エンドポイントの一覧と設定確認
  • モデル配置前のエンドポイント作成
  • 表示名やラベルを変更したいとき
google/skills2.2万2026年10月28日 更新

agent-platform-eval-flywheel

Google 公式無料概要あり

Google Cloud上のAIモデルやエージェントを評価し、テストデータ作成、採点指標の選定、失敗分析、修正前後の比較を通じて品質改善を支援します。

  • 会話ログから評価データを作りたいとき
  • 模擬ユーザーのシナリオ生成
  • 回答やツール利用の評価指標選び
google/skills2.2万2026年10月28日 更新

agent-platform-inference

Google 公式無料概要あり

Google CloudのAgent PlatformでGeminiやLlamaなどに質問を送り、回答を得るための認証・接続コード作成と実行、エラー対処を支援するスキル。

  • Geminiにテスト用の質問を送りたいとき
  • OpenAI SDKでLlama等を呼びたいとき
  • 既存の接続先を呼ぶコードの作成
google/skills2.2万2026年10月28日 更新

agent-platform-migrate-from-ai-studio

Google 公式無料概要あり

Google AI StudioのGemini APIをGoogle Cloudへ移す手順を案内し、認証・権限・課金の設定からSDKやエージェントの接続変更まで支援します。

  • GeminiアプリをGoogle Cloudへ移したいとき
  • CloudクレジットをAPI利用に充てたいとき
  • エージェントの認証と課金を統合したいとき
google/skills2.2万2026年10月28日 更新

google のスキルをすべて見る

このスキルの問題を報告する