Google Developer Knowledge
Retrieve current, source-backed official Google developer documentation for Google Cloud, Firebase, Android, Google AI, Gemini CLI, Flutter, Go, Maps, Web, T...
wei
@wei840222
What This Skill Does
Retrieves current, source-backed official Google developer documentation for Google Cloud, Firebase, Android, Google AI, Gemini CLI, Flutter, Go, Maps, Web, TensorFlow, and related products. Uses mcporter to discover and call a Google Developer Knowledge-capable MCP server for search, full-document retrieval, and grounded answers.
Replaces manually browsing multiple Google developer doc sites and outdated cached pages by providing a single, verified, source-backed documentation retrieval workflow.
When to Use It
- Look up current pricing and quota details for a Google Cloud service
- Find official setup steps for migrating a Firebase project to a new region
- Get the latest API reference for Android development
- Troubleshoot a Google AI or Gemini CLI integration with official docs
- Compare Cloud Run and Cloud Functions for a new HTTP microservice using grounded documentation
- Retrieve full-page context for a version-sensitive Google developer product change
Install
$ openclaw skills install @wei840222/google-developer-knowledgeGoogle Developer Knowledge โ Official Google Docs Skill
Use this skill when the task needs current official Google developer documentation for Google Cloud, Firebase, Android, Google AI, Gemini CLI, Flutter, Go, Maps, Web, TensorFlow, or other products indexed by the Developer Knowledge corpus.
Core Rule
Use mcporter capability discovery from references/mcporter-workflow.md. Select only a healthy server exposing Google Developer Knowledge search, full-document retrieval, and grounded-answer capabilities. Do not assume a server alias, URL, tool-name prefix, authentication arrangement, or native tool wrapper.
Tool Choice
- Start with
search_documentsfor most research. It returns official documentation chunks withparent,id, andcontent. - Use
get_documentswith returnedparentdocument names when snippets are insufficient or citations need full-page context. - Use
answer_queryfor a quick grounded synthesis only when its references are sufficient. If it returns 429, times out, or lacks adequate references, fall back tosearch_documentsplusget_documents.
Decision Checkpoints
- ๐ด CHECKPOINT โ Capability discovery: Parse the selected route returned by discovery into
gdk_server,gdk_search_tool,gdk_get_tool, andgdk_answer_tool. If discovery finds no compatible server, record the reason and use REST only withDEVELOPERKNOWLEDGE_API_KEYor OAuth credentials plus a quota project; otherwise use an official-document fallback. - ๐ด CHECKPOINT โ Evidence sufficiency: For version-sensitive, pricing, quota, security, or migration claims, retrieve the relevant full document with
get_documents. A search snippet or ananswer_queryresponse alone is not enough. - ๐ด CHECKPOINT โ Retrieval budget: Retrieve no more than three parent documents by default. Ask before broad extraction, offline use, or a larger document set.
- ๐ด CHECKPOINT โ Output trace: State the selected server or fallback, the retrieval method, and source URLs or document URIs. If the needed detail is absent, mark it unverified rather than infer it.
Quick Commands
# Run capability discovery from references/mcporter-workflow.md first.
gdk_route='<selected-route-json>'
gdk_server="$(jq -r '.server' <<<"$gdk_route")"
gdk_search_tool="$(jq -r '.tools.search' <<<"$gdk_route")"
gdk_get_tool="$(jq -r '.tools.get' <<<"$gdk_route")"
gdk_answer_tool="$(jq -r '.tools.answer' <<<"$gdk_route")"
mcporter call "$gdk_server.$gdk_search_tool" \
--args '{"query":"Cloud Run minimum instances concurrency pricing"}' \
--output json
mcporter call "$gdk_server.$gdk_get_tool" \
--args '{"names":["documents/docs.cloud.google.com/run/docs/about-concurrency"]}' \
--output json
mcporter call "$gdk_server.$gdk_answer_tool" \
--args '{"query":"Compare Cloud Run and Cloud Functions for a new HTTP microservice."}' \
--output json
Workflow
- Define one focused documentation question.
- Pass Capability discovery and retain the returned route and tool names for the task.
- Search with
search_documents. - Pull the best
parentdocuments withget_documentswhen evidence requires full context. - Pass Evidence sufficiency, then synthesize only from retrieved evidence.
- Pass Retrieval budget and Output trace before responding.
Coverage and Limits
- Corpus covers public pages from Google developer domains such as
cloud.google.com,docs.cloud.google.com,firebase.google.com,developer.android.com,ai.google.dev,developers.google.com,web.dev,go.dev,docs.flutter.dev, andwww.tensorflow.org. - Some generated API reference prefixes are excluded from the corpus, especially language-specific Google Cloud reference directories.
- Data freshness target is usually 24-48 hours after documentation publication.
- Full documents are Markdown and can be large; retrieve only the pages needed for the answer.
Detailed Reference
Use references/mcporter-workflow.md for command patterns, field shapes, troubleshooting, and source-backed answer patterns.
Use references/api-fallback.md only after capability discovery fails, when validating a client integration, or when the task specifically asks for direct Developer Knowledge REST API calls. Never attempt REST without an API key or OAuth credentials plus a quota project.
Retrieval Record
For answers that depend on current Google behavior, include a concise record:
method:mcp,rest-api-key,rest-oauth, orofficial-document.server: selected MCP server, ornonewhen fallback was used.fallback_reason:none, discovery failure, rate limit, authentication failure, or corpus gap.sources: official URLs or document URIs used.verification:full-document,search-snippet-only, orunverified.
Anti-Patterns
| Do not | Why | Correct action |
|---|---|---|
| Hardcode a server alias or endpoint | Server names and authentication differ between runtimes. | Use capability discovery and retain the selected gdk_server for the task. |
| Treat a search snippet or unreferenced synthesis as proof | It can omit constraints or be stale for consequential claims. | Fetch the parent document and cite its URI before making the claim. |
| Retry timeout, rate-limit, or authentication errors without a bound | It hides an integration failure and can amplify requests. | Follow the Failure Matrix: one bounded 429 retry, then fallback or stop. |
| Call REST without an API key or OAuth token plus quota project | It creates a predictable authentication or quota failure. | Use REST only after an authentication preflight; otherwise use official Google documentation. |
| Claim a grounded answer after MCP or retrieval fails | The result is no longer evidence-backed. | State the fallback reason and mark unavailable details as unverified. |
| Send sensitive, proprietary, or credential material in documentation queries | Queries can leave the current environment. | Use product names, public error text, and minimal non-sensitive context only. |
Official Documents
These are the official documents this skill was built from:
https://developers.google.com/knowledge/apihttps://developers.google.com/knowledge/quickstarthttps://developers.google.com/knowledge/howtohttps://developers.google.com/knowledge/answer-queryhttps://developers.google.com/knowledge/mcphttps://developers.google.com/knowledge/reference/mcphttps://developers.google.com/knowledge/reference/mcp/tools_list/search_documentshttps://developers.google.com/knowledge/reference/mcp/tools_list/answer_queryhttps://developers.google.com/knowledge/reference/mcp/tools_list/get_documentshttps://developers.google.com/knowledge/reference/corpus-reference
Top skills in this category
Marketing Mode
@thesethroseMarketing Mode combines 23 comprehensive marketing skills covering strategy, psychology, content, SEO, conversion optimization, and paid growth. Use when users need marketing strategy, copywriting, SEO help, conversion optimization, paid advertising, or any marketing tactic.
Marketing Skills
@jchopard69Access 23 marketing modules offering checklists, frameworks, and ready-to-use deliverables for CRO, SEO, copywriting, analytics, launches, ads, and social me...
Prompt Guard
@seojoonkim650+ pattern AI agent security defense covering prompt injection, supply chain injection, memory poisoning, action gate bypass, unicode steganography, cascad...
Computer Use
@ram-raghav-sFull desktop computer use for headless Linux servers. Xvfb + XFCE virtual desktop with xdotool automation. 17 actions (click, type, scroll, screenshot, drag,...
YouTube Transcript
@xthezealotFetch and summarize YouTube video transcripts. Use when asked to summarize, transcribe, or extract content from YouTube videos. Handles transcript fetching via residential IP proxy to bypass YouTube's cloud IP blocks.