Skip to main content
Agent skill

Retrieving Developer Knowledge

Searches, retrieves, and synthesizes official Google developer documentation across Google Cloud, AI/Gemini, Android, Chrome, Web, Flutter, Go, Firebase, and other Google developer platforms. Integrates with the Developer Knowledge MCP server (searchdocuments, getdocuments, answerquery) or the Developer Knowledge REST API fallback. Use when searching for gcloud CLI commands, API syntax, IAM permissions, official documentation, architectural comparisons, or product choice overviews. Don't use for local filesystem lookups or non-Google documentation.

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Retrieving Developer Knowledge when an AI agent needs a reusable SKILL.md workflow for this job: Searches, retrieves, and synthesizes official Google developer documentation across Google Cloud, AI/Gemini, Android, Chrome, Web, Flutter, Go, Firebase, and other Google developer platforms. Integrates with the Developer Knowledge MCP server (searchdocuments, getdocuments, answerquery) or the Developer Knowledge REST API fallback. Use when searching for gcloud CLI commands, API syntax, IAM permissions, official documentation, architectural comparisons, or product choice overviews. Don't use for local filesystem lookups or non-Google documentation.

When not to use it

Skip Retrieving Developer Knowledge when the task is outside the coding category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/retrieving-developer-knowledge
# Save SKILL.md to ~/.claude/skills/retrieving-developer-knowledge/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/retrieving-developer-knowledge/SKILL.md. For a shared project, commit the folder at .claude/skills/retrieving-developer-knowledge/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Retrieving Developer Knowledge skill for this task: Direct Retrieval: When answering a technical question, execute a single documentation lookup directly within your current conversation context (do not delegate retrieval to subagents):. Confirm any required inputs or files before making changes.

What supports this recommendation?

83/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
92/100
Practical detail
92/100
Source evidence
63/100
Source signals
72/100
Model compatibility evidence
  • claudeuntested

    No model-specific signal or recorded compatibility test was found.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiinferred

    The skill text mentions Gemini or a closely associated term. This is keyword-derived compatibility; no hands-on model test is recorded.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# Google Developer Knowledge

The Developer Knowledge skill provides access to official Google developer documentation across Google Cloud, AI/ML (ai.google.dev, ADK, TensorFlow), Android, Chrome, Web, Flutter, Go, Firebase, and other Google developer platforms via the Developer Knowledge MCP server or REST API fallback.

## Workflow

1. **Direct Retrieval**: When answering a technical question, execute a single documentation lookup directly within your current conversation context (do not delegate retrieval to subagents): - **If MCP tools are present in your environment**: Call `answer_query` (for conceptual guides/workflows) or `search_documents` (for CLI flags/syntax). - **If MCP tools are not present**: Execute a REST API request via `curl` against `https://developerknowledge.googleapis.com/v1`. - **A declared server is not always a connected server.** Some clients cannot complete the MCP handshake with this server and expose no `answer_query`, `search_documents` or `get_documents` tool at all, even though the plugin declares one. Treat their absence as normal and use the REST fallback below. 2. **Confirm the lookup succeeded before using it**: A response that arrives is not automatically an answer. `PERMISSION_DENIED`, `UNAUTHENTICATED`, HTTP 401 or 403, an empty result set, or any error payload is a FAILED lookup even when the tool itself reported no error. On a failed lookup, do not answer as though it had succeeded. - **Rate Limits (HTTP 429 / Out of Quota)**: If calling `answer_query` or the REST `:answerQuery` endpoint returns an HTTP 429 (`RESOURCE_EXHAUSTED` / out of quota error), this indicates server-side generative QA quota exhaustion. Do not treat this as an unrecoverable failure or fall back to ungrounded memory; immediately retry using the search endpoint (`search_documents` for MCP, or `documents:searchDocumentChunks` for REST) with 2–5 focused keywords from the query. - **Persistent Failures**: Try the other transport once, and if that also fails, state plainly in your reply to the user that you could not reach Developer Knowledge and are answering without it. Presenting recalled documentation as a retrieved result is the worst available outcome, because nothing in the reply distinguishes it from a real lookup. 3. **Immediate & Complete Solution Output**: Immediately upon receiving the documentation response, output the complete, self-contained, and executable technical solution (commands with all required flags and placeholders, YAML/JSON configurations, or code snippets) directly in your response text.

## Tool Selection & Usage

Choose the appropriate tool based on availability in your runtime environment:

### 1. Developer Knowledge MCP Tools (Preferred) When MCP tools are present in your active tool definitions: - **`answer_query(query="...")`**: Use for conceptual guides, architectural comparisons, product choice overviews, and multi-step workflows. This tool has limited quota. If you get a 429 out of quota error, use `search_documents` instead. - **`search_documents(query="...")`**: Use for granular CLI flags, exact syntax, parameter names, and IAM permissions (`service.resource.verb`). Use 2–5 focused keywords (e.g., `cloud run filestore nfs mount gcloud`) rather than full conversational sentences. - **`get_documents(names=["documents/{uri_without_scheme}"])`**: Fetch full documentation pages by resource name (e.g. `names: ["documents/docs.cloud.google.com/run/docs/overview/what-is-cloud-run"]`).

### 2. REST API Fallback When the MCP tools are absent, query the Developer Knowledge REST API (`https://developerknowledge.googleapis.com/v1`) with `curl`. Two credentials work, and you should try them in this order.

**An existing Google credential, preferred.** If `gcloud` is authenticated, pass a bearer token and your quota project. Nothing needs to be installed or configured:

```bash curl -s -X POST "https://developerknowledge.googleapis.com/v1:answerQuery" \ -H "Authorization: Bearer $(gcloud auth print-access-token)" \ -H "X-Goog-User-Project: $(gcloud config get-value project 2>/dev/null)" \ -H "Content-Type: application/json" \ -d "{\"query\": \"How do I configure public read access on Cloud Storage?\"}" ```

If that fails to authenticate, on a 401, a 403, or any other credential error, the account has a token the API will not accept. Substitute `gcloud auth application-default print-access-token` for `gcloud auth print-access-token` in the command above and try again. Which credential the API accepts depends on how the environment was authenticated, so treat an auth error here as a reason to try the application-default credential rather than as a failed lookup.

**An API key, if one is configured.** Where `DEVELOPERKNOWLEDGE_API_KEY` is set in the environment, pass it as a `key` query parameter instead of an `Authorization` header. If neither credential is available, an API key is the supported path for this client: enable the API and create one by following the [Developer Knowledge quickstart](https://developers.google.com/knowledge/quickstart), then export it as `DEVELOPERKNOWLEDGE_API_KEY`. Say that you need a credential rather than answering without a lookup. The remaining examples in this section use that form: - **Answer Query**: ```bash curl -s -X POST "https://developerknowledge.googleapis.com/v1:answerQuery?key=${DEVELOPERKNOWLEDGE_API_KEY}" \ -H "Content-Type: application/json" \ -d '{"query": "How do I configure public read access on Cloud Storage?"}' ``` This endpoint has limited quota. If you get a 429 out of quota error, use `documents:searchDocumentChunks` instead. - **Search Document Chunks** (use 2–5 focused keywords): ```bash curl -s "https://developerknowledge.googleapis.com/v1/documents:searchDocumentChunks?query=gcloud+logging+metrics+create&key=${DEVELOPERKNOWLEDGE_API_KEY}" ``` - **Get Document**: ```bash curl -s "https://developerknowledge.googleapis.com/v1/documents/docs.cloud.google.com/run/docs/overview/what-is-cloud-run?key=${DEVELOPERKNOWLEDGE_API_KEY}" ``` - **Batch Get Documents** (a `GET`, with one `names` parameter per document and no request body; up to 20 per call): ```bash curl -s -G "https://developerknowledge.googleapis.com/v1/documents:batchGet" \ --data-urlencode "names=documents/docs.cloud.google.com/run/docs/overview/what-is-cloud-run" \ --data-urlencode "names=documents/docs.cloud.google.com/storage/docs/creating-buckets" \ --data-urlencode "key=${DEVELOPERKNOWLEDGE_API_KEY}" ```

## Synthesis & Output Guidelines

1. **Grounding in Official Documentation**: Ground all solutions directly in retrieved documentation. Official documentation conventions have absolute precedence over memorized defaults. 2. **Exact Parameter Formatting**: Format CLI flags, composite keys (e.g. `location=IP:PATH`), and IAM permission strings according to official Google specifications. 3. **Complete Solutions in Final Response**: Always output the full, self-contained, executable technical solution (commands, configurations, or code snippets) with clear standard placeholders (e.g. `PROJECT_ID`, `SERVICE_NAME`, `REGION`) directly in your final message, even if previously referenced during internal planning.

## References

- [MCP Usage & Tool Details](references/mcp-usage.md) - [REST API Fallback Guide](references/api-fallback.md) - [Supported Domains & Scoping](references/supported-domains.md)

Intended uses

  • Direct Retrieval: When answering a technical question, execute a single documentation lookup directly within your current conversation context (do not delegate retrieval to subagents):
  • If MCP tools are present in your environment: Call answerquery (for conceptual guides/workflows) or searchdocuments (for CLI flags/syntax).
  • If MCP tools are not present: Execute a REST API request via curl against https://developerknowledge.googleapis.com/v1.
  • A declared server is not always a connected server. Some clients cannot complete the MCP handshake with this server and expose no answerquery, searchdocuments or getdocuments tool at all, even though the plugin declares one. Treat their absence as normal and use the REST fallback below.
  • Confirm the lookup succeeded before using it: A response that arrives is not automatically an answer. PERMISSIONDENIED, UNAUTHENTICATED, HTTP 401 or 403, an empty result set, or any error payload is a FAILED lookup even when the tool itself reported no error. On a failed lookup, do not answer as though it had succeeded.
  • Rate Limits (HTTP 429 / Out of Quota): If calling answerquery or the REST :answerQuery endpoint returns an HTTP 429 (RESOURCEEXHAUSTED / out of quota error), this indicates server-side generative QA quota exhaustion. Do not treat this as an unrecoverable failure or fall back to ungrounded memory; immediately retry using the search endpoint (searchdocuments for MCP, or documents:searchDocumentChunks for REST) with 2–5 focused keywords from the query.
  • Persistent Failures: Try the other transport once, and if that also fails, state plainly in your reply to the user that you could not reach Developer Knowledge and are answering without it. Presenting recalled documentation as a retrieved result is the worst available outcome, because nothing in the reply distinguishes it from a real lookup.
  • Immediate & Complete Solution Output: Immediately upon receiving the documentation response, output the complete, self-contained, and executable technical solution (commands with all required flags and placeholders, YAML/JSON configurations, or code snippets) directly in your response text.

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.