gemini-docs-skill

USE THIS SKILL for any question about Google Gemini API — models, function calling, structured output, Live API, embeddings, pricing, or SDKs. This skill has complete local documentation — do not use

SKILL.md

GitHub

Gemini API Documentation Skill

Intelligent documentation access for Google's Gemini API.

Path Resolution

Use ${CLAUDE_SKILL_DIR} to reference this skill's directory — it resolves automatically whether the skill is symlinked or local.

Overview

This skill provides smart-routed access to comprehensive Gemini API documentation:

  • ~64 pages of official documentation from ai.google.dev
  • Flat structure — all pages at the same level (no subsections)
  • 3-tier architecture for optimal context usage and fast lookups

Documentation Update Requests

When user requests to update documentation (trigger: "update gemini documentation"):

Update All Documentation

"${CLAUDE_SKILL_DIR}/scripts/update-docs.sh"

What Gets Updated:

  1. Quick Reference - Tier 1 — Key features from priority pages
  2. Section Index - Tier 2 — Alphabetical page summaries
  3. Full Documentation - Tier 3 — Complete documentation with grep markers

Duration: ~15 seconds (low rate limit, fast .md.txt endpoints)

Error Handling:

  • Script logs each page fetch status
  • Failed pages are skipped with warnings
  • Aborts if <10 pages fetched (critical failure)

After Update:

  1. Confirm all tiers generated
  2. Report page count and any failures
  3. New documentation immediately available (hot-reload)

3-Tier Routing

Tier 1: Quick Reference

  • File: references/quick-reference.md
  • Covers: Quickstart, models, text generation, function calling, structured output, Live API, pricing, SDKs
  • Use when: Quick answers, common features, getting started questions

Example queries:

  • "How do I get started with Gemini?"
  • "What models are available?"
  • "How does function calling work in Gemini?"
  • "What's the pricing for Gemini API?"

Access pattern:

# Read the quick reference directly
cat references/quick-reference.md

Tier 2: Section Index

  • File: references/section-index.md
  • Covers: All pages with titles and descriptions, sorted alphabetically
  • Use when: Finding specific topics, understanding documentation scope

Example queries:

  • "What topics does the Gemini docs cover?"
  • "Is there documentation about caching?"
  • "Where are the OAuth docs?"

Access pattern:

# Find a specific topic
grep -i "caching" references/section-index.md

# List all available pages
grep "^###" references/section-index.md

Tier 3: Full Documentation

  • File: references/full-docs.txt
  • Covers: Complete documentation with grep markers per page
  • Use when: Detailed information, specific page content, deep dives

IMPORTANT: Never load entire file. Use grep markers for targeted extraction.

Example queries:

  • "Show me the full function calling docs"
  • "How do I use structured output with Gemini?"
  • "What are the safety settings options?"
  • "How does the Live API session management work?"

Access patterns:

# Find and extract a specific page
grep -A 100 "^PAGE: /function-calling" references/full-docs.txt

# Search for a keyword across all docs
grep -n "function_declarations" references/full-docs.txt

# Extract complete page between separators
sed -n '/^PAGE: \/structured-output$/,/^═\{80\}$/p' references/full-docs.txt

# List all pages
grep "^PAGE:" references/full-docs.txt

Coverage

Topic Path Description
Quickstart /quickstart Getting started with the Gemini API
Models /models All available Gemini models
Text generation /text-generation Chat and text generation
Function calling /function-calling Tool use / function calling
Structured output /structured-output JSON schema output
Live API /live Real-time streaming API
Pricing /pricing Model pricing and rate limits
Libraries / SDKs /libraries Official client libraries
Audio /audio Audio understanding
Image understanding /image-understanding Vision capabilities
Image generation /image-generation Generating images
Video understanding /video-understanding Video analysis
Video generation /video Veo video generation
Code execution /code-execution Sandboxed code execution
Embeddings /embeddings Text embeddings
Caching /caching Context caching
Long context /long-context 1M+ token context
Thinking /thinking Chain-of-thought reasoning
Grounding /google-search Google Search grounding
Safety settings /safety-settings Content safety configuration
Fine-tuning /model-tuning Model fine-tuning
OpenAI compat /openai OpenAI API compatibility
Deep Research /deep-research Multi-step research tasks
Computer Use /computer-use Desktop automation
Coding Agents /coding-agents Coding agent patterns

Query Routing Examples

"How do I get started with Gemini API?"
-> Tier 1: quick-reference.md (Quickstart section)

"What models are available?"
-> Tier 1: quick-reference.md (Models section)

"Is there documentation on video generation?"
-> Tier 2: section-index.md (find the page)

"How does function calling work in detail?"
-> Tier 3: grep "^PAGE: /function-calling" full-docs.txt

"What safety settings are available?"
-> Tier 3: grep "^PAGE: /safety-settings" full-docs.txt

"Compare Gemini pricing tiers"
-> Tier 1 first, then Tier 3 for details: grep "^PAGE: /pricing" full-docs.txt

Response Format

Always cite tier and source:

  • "From Quick Reference: ..."
  • "From Section Index: ..."
  • "From Full Docs (/function-calling): ..."

Provide context:

Best Practices

DO

  • Start with Tier 1 (quick-reference.md) for common questions
  • Use Tier 2 (section-index.md) to find relevant pages
  • Use targeted grep/sed for Tier 3 (never load full file)
  • Cite sources in responses

DON'T

  • Load entire full-docs.txt file
  • Answer from memory — always check current docs
  • Skip tier routing (always start with appropriate tier)
  • Use web search when this skill is installed

File Locations

gemini-docs-skill/
├── SKILL.md
├── docs-config.yaml
├── references/
│   ├── quick-reference.md     # Tier 1
│   ├── section-index.md       # Tier 2
│   └── full-docs.txt          # Tier 3
└── scripts/
    └── update-docs.sh         # Wrapper (calls shared doc-builder)

Maintenance

Update Schedule

Frequency Reason
Monthly Google updates Gemini API regularly
On request When user asks about features not in docs
After major Gemini updates New models, API changes

Update Command

"${CLAUDE_SKILL_DIR}/scripts/update-docs.sh"

Version

Skill Version: 1.0.0 Documentation Source: https://ai.google.dev/gemini-api/docs Last Updated: 2026-03-01


Remember: Check Tier 1 first -> Check Tier 2 to find topics -> Use Tier 3 for deep details