manage-headers

作者: microsoft

检查并配置Power Pages站点发送给浏览器的安全标头——内容安全策略、框架与点击劫持保护、跨源……

npx skills add https://github.com/microsoft/power-platform-skills --skill manage-headers

Plugin check: Run node "${PLUGIN_ROOT}/scripts/check-version.js" — if it outputs a message, show it to the user before proceeding.

Manage Headers

Inspect and configure the HTTP security headers for a Power Pages site. Headers are configured as HTTP/* site settings stored in .powerpages-site/site-settings/ YAML files.

Initial request: $ARGUMENTS

Gotchas

  • Site settings are YAML files. Each header is a separate .yml file in .powerpages-site/site-settings/. The file name uses - instead of / (e.g., HTTP/X-Frame-Options → http-x-frame-options.sitesetting.yml).
  • Absent = no header. When a site setting is absent, the runtime omits that header entirely (except CSP on new sites — see headers-reference.md).
  • HSTS and Cache-Control are platform-managed. Do not try to set HTTP/Strict-Transport-Security — the runtime does not recognize it and the setting has no effect.
  • Maker-mode bypasses headers. Requests from Power Pages Studio skip all HTTP/* header emission. Verify headers in an incognito tab, not the studio preview.
  • CSP is pass-through. The runtime emits the value verbatim — it does NOT merge runtime sources automatically. The CSP MUST include Power Pages runtime hosts or the site breaks.
  • CSP nonce. When script-src contains 'nonce', the runtime replaces it per-request with 'nonce-<random>' and auto-hashes inline event handlers. Scripts created dynamically via document.createElement do NOT receive the nonce.
  • SameSite=None requires HTTPS. The runtime sets Secure on every cookie over HTTPS automatically.
  • CORS * is auto-specialized. The runtime replaces * per-request with the specific requesting Origin — the browser sees a single-origin header, not a wildcard.

Workflow

  1. Prerequisites — Locate project, confirm site-settings directory exists
  2. Inspect current headers — Read site-setting YAML files, identify configured and missing headers
  3. Assess and plan — Identify gaps, present recommendations
  4. Apply changes — Edit existing settings or create new ones
  5. Summarize — Present results, record usage, offer follow-ups

Task Tracking

Create tasks in four groups. Mark each in_progress when starting, completed when done.

GroupWhen to createTasks
1At startCheck prerequisites
2After prerequisites passInspect current headers · Assess and plan (skip "Assess and plan" in review mode)
3After user approves changesApply changes (skip in review mode OR if no changes were accepted)
4After apply or assessSummarize (always)

1. Prerequisites

1.1 Locate the project, detect review mode

Use Glob to find **/powerpages.config.json. If $ARGUMENTS contains --review <out-dir>, remember the output directory — Steps 3–4 are skipped and Step 5 writes JSON only.

1.2 Verify site-settings directory

Check that .powerpages-site/site-settings/ exists. If not, the site has not been deployed yet — tell the user and recommend /deploy-site. Stop.


2. Inspect current headers

Use Glob to find all *.yml files in .powerpages-site/site-settings/. Use Read to read each file and extract the name and value fields. Identify all settings with an HTTP/ prefix — these are the configured headers.

Compare against the recognized header catalogue in references/headers-reference.md. For each header in the catalogue:

  • Present — record its current value.
  • Missing — record it as absent and note the recommended value from headers-reference.md.

For CSP specifically: if HTTP/Content-Security-Policy is present, scan the project's source files using Glob + Read to find external URLs and check whether they are covered by the policy. Identify the site's cloud environment via pac auth who to determine the correct Power Pages runtime host (see headers-reference.md § "Power-Pages-runtime sources a CSP must allow").


3. Assess and plan

Skip in review mode.

MUST use plain language only. Never lead with words like CSP, CORS, HSTS, or MIME sniffing — explain using everyday language:

Header conceptPlain-language name
Content-Security-Policy"which scripts and resources the browser is allowed to load"
X-Frame-Options / frame-ancestors"whether other websites can put your site inside a frame"
X-Content-Type-Options"stop the browser from guessing file types"
CORS headers"which other websites can call your site's data"
SameSite cookies"when the browser sends your sign-in cookie"

Default approach

Read references/headers-reference.md for recommended values and guidance. Present the most important gaps first — headers that are missing or misconfigured relative to the recommended values.

🚦 Gate (plan · manage-headers:3.per-finding): Per-finding loop — for each header gap, prompt accept / customize / skip. Fires PER FINDING in the loop; skipped findings leave the header at its current value, accepted/customized findings get an Edit / create-script call in Phase 4.

Trigger: Phase 3 entry has tallied header gaps against references/headers-reference.md. Why we ask: Auto-accepting can apply CSP/CORS values that break the site (legitimate scripts blocked, third-party widgets refused); auto-skipping leaves the site missing important headers. Cancel leaves: Nothing — Phase 4's Edit / create-script call only fires on accepted findings.

For each finding, present via AskUserQuestion:

  • A plain-language explanation of why the change matters
  • The recommended value
  • Options: accept the recommendation, customize, or skip

Do NOT present all headers at once — present the important gaps first. For headers already set to recommended values, mention them in the summary without requiring action.

CSP composition

When the user needs a CSP (missing or incomplete), compose one using:

  1. The starter template from headers-reference.md
  2. The correct cloud-specific runtime host
  3. External URLs discovered from the project's source files — scan ALL source files, templates, scripts, etc.
  4. The 'nonce' keyword for inline scripts

When reviewing an existing CSP, validate:

  • All external URLs actually loaded by the site are covered in the policy
  • Runtime hosts are included for the site's cloud

Present the composed or corrected CSP for review. Recommend starting in report-only mode (HTTP/Content-Security-Policy-Report-Only) before enforcing.


4. Apply changes

Skip in review mode.

For existing settings: use Edit on the YAML file directly — change the value field.

For new settings: use the shared create script:

node "${PLUGIN_ROOT}/scripts/create-site-setting.js" \
  --projectRoot "<PROJECT_ROOT>" \
  --name "<setting-name>" \
  --value "<value>" \
  --description "<description>"

See references/commands.md for details.

After all changes are applied, offer to deploy: "Ready to deploy these changes? They take effect after the next deploy." If yes, invoke /deploy-site.


5. Summarize

5.1 Review mode

First, read the configured HTTP/* site settings (from Step 2 — you already have them). Then write <REVIEW_DIR>/header-annotations.json with a plain-language description for each header and, when the configured value has a genuine issue (missing critical directive, weak value), a suggested fix. The transform script no longer hardcodes header descriptions — they come from you.

{
  "headers": {
    "HTTP/<HeaderName>": { "description": "What this header does, in plain language.", "fix": "Optional fix if the configured value has a genuine issue." }
  }
}

Use references/headers-reference.md for authoritative descriptions and validation rules. Surface a fix only when the value has a real problem — do not editorialize on every header.

Then run the transform:

node "${PLUGIN_ROOT}/skills/manage-headers/scripts/transform-headers.js" \
  --projectRoot "<PROJECT_ROOT>" \
  --annotations "<REVIEW_DIR>/header-annotations.json"

Write the stdout to <REVIEW_DIR>/manage-headers.json and stop. The transform emits { status, findings, details }; the orchestrating skill handles presentation.

5.2 Present summary

Skip in review mode.

Plain-language summary: what was changed, what gaps remain, and what is already well-configured.

5.3 Record skill usage

Reference: ${PLUGIN_ROOT}/references/skill-tracking-reference.md

Use --skillName "ManageHeaders".

5.4 Offer follow-ups

If a natural follow-up exists based on findings, suggest it. If no meaningful follow-up exists, end the skill.


Constraints

  • Plain language — MUST NOT use technical jargon with the user. Explain header names using everyday language.
  • headers-reference.md is the source of truth — recommended values and the recognized header catalogue live there. Read it before assessing.
  • Context-aware interactions — every recommendation MUST reflect the site's actual configuration and usage:
    • Read the site's source files, integrations, and auth setup before recommending any value.
    • Never recommend a value without verifying it will not break the site's functionality (see the "Context to verify" column in headers-reference.md).
    • Acknowledge existing values when proposing changes — they may be intentional.
    • For CSP, reference actual external URLs found in the project's source files.
    • For CORS, verify the site's actual cross-origin consumers before scoping.
    • For Cross-Origin-Opener-Policy, verify whether the site uses popup-based auth.
    • For Cross-Origin-Resource-Policy, verify whether the site hosts Azure AD B2C custom login pages, is embedded cross-origin, or has integrations that load its resources. Leave absent or use cross-origin when unsure.
  • Preview is for change review only — include preview only on options that modify a setting value. Do not add to informational choices.
  • Recommendations MUST NOT break the site — before recommending a value, consider whether it would block resources the site actually uses. For CSP, always verify that runtime sources and project external URLs are included. For CORS, verify the site's actual cross-origin needs. When unsure, recommend report-only mode first.
  • NEVER recommend broadening an existing policy — if the user already has a tight CSP, CORS scope, or restrictive header value, do not suggest making it less restrictive. A working tight policy is better than a broad one. Never recommend https: wildcards in CSP directives — list specific hosts instead.
  • Deploy after changes — header changes only take effect after deploying. Always offer /deploy-site after applying changes.

References

  • references/headers-reference.md — recognized header catalogue, recommended values, CSP composition rules, runtime sources. Read before Step 2 (inspect) and Step 3 (assess) in interactive mode.
  • references/commands.md — shared create-site-setting.js usage. Read at Step 4 (apply) when creating new settings.

来自 microsoft 的更多技能

oss-growth
microsoft
OSS增长黑客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。在创建使用AzureAIAgentsProvider的持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应时使用。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上设置AI Runway——从裸集群到运行模型。涵盖集群验证、控制器安装、GPU评估、提供商设置和首次部署。适用场景:“设置AI Runway”、“接入AKS集群”、“安装AI Runway”、“airunway设置”、“将模型部署到AKS”、“在AKS上进行GPU推理”、“在AKS上配置KAITO”、“在AKS上运行LLM”、“在AKS上使用vLLM”、“在AKS上设置模型服务”、“AI Runway控制器”。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights对Web应用进行插桩的指南。提供遥测模式、SDK设置和配置参考。适用场景:如何对应用进行插桩、App Insights SDK、遥测模式、什么是App Insights、Application Insights指南、插桩示例、APM最佳实践。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)为浏览器/Web应用添加检测。用于真实用户监控(RUM)——页面视图、点击、AJAX/fetch依赖项、异常、自定义事件,以及与后端OpenTelemetry追踪关联的浏览器端GenAI代理追踪。涵盖SDK加载器脚本和npm设置、框架扩展(React、React Native、Angular)、点击分析、遥测初始化器,以及从浏览器发出的代理/工具/模型跨度所遵循的OTel GenAI语义约定。
devops
azure-ai-anomalydetector-java
microsoft
使用适用于 Java 的 Azure AI 异常检测器 SDK 构建异常检测应用程序。在实现单变量/多变量异常检测、时间序列分析或 AI 驱动的监控时使用。
development
azure-ai-language-conversations-py
microsoft
使用azure-ai-language-conversations Python SDK实现对话语言理解(CLU)。当使用ConversationAnalysisClient分析对话意图和实体、构建NLP功能或将语言理解集成到应用程序中时使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用于机器学习工作区、作业、模型、数据集、计算资源和管道。 触发词:“azure-ai-ml”、“MLClient”、“工作区”、“模型注册表”、“训练作业”、“数据集”。
development