google-patents-search

Google Patents Search

Search Google Patents results through Frevana.

Purpose

This skill is for finding Google Patents results by patent query.

Inputs:

• q
• optional page
• optional num
• optional language
• optional status

Output:

• validated response JSON with Google Patents results
• the same validated JSON saved to a local file on every successful run

This skill validates that the response is JSON, saves it to disk, and returns it unchanged on stdout. Do not rewrite or reshape the returned data unless the user explicitly asks for a transformation.

What This Skill Needs

• user-provided q
• optional page
• optional num
• optional language
• optional status
• FREVANA_TOKEN in the environment variables, or an explicit --token override for the current run
• curl
• bash
• python3

Execution Order

Use this flow so the request stays simple and reliable:

1. Confirm the user has provided q or an equivalent patent search query.
2. Prefer the script over ad hoc curl commands.
3. Let the script read FREVANA_TOKEN first.
4. In interactive shell usage, if FREVANA_TOKEN is missing, the script may prompt for it.
5. In non-interactive or agent workflows, fail fast if the token is missing and tell the user to set FREVANA_TOKEN or pass --token explicitly.
6. Run the script once. It prints the validated JSON to stdout and saves the same JSON to a file.
7. Use the saved JSON file for any follow-up parsing or summarization instead of calling the search API again.
8. Return the validated response JSON, or summarize the most relevant patent results if the user does not need the full payload.

Commands

Basic patents search

bash <skill-path>/scripts/search_google_patents.sh \
  --q "(Coffee)"

--query is accepted as an alias for --q.

Search with status, language, page, and result count

bash <skill-path>/scripts/search_google_patents.sh \
  --q "(Coffee)" \
  --status GRANT \
  --language en \
  --page 0 \
  --num 10

Save response JSON to a specific file

bash <skill-path>/scripts/search_google_patents.sh \
  --q "(Coffee)" \
  --output ./out/google-patents-search-result.json

Token override for the current run

bash <skill-path>/scripts/search_google_patents.sh \
  --q "(Coffee)" \
  --token "your bearer token"

Fixed Request Shape

The script sends this payload shape, omitting optional fields that were not provided:

{
  "q": "(Coffee)",
  "page": 0,
  "num": 10,
  "language": "en",
  "status": "GRANT"
}

Only q is required. Do not invent page, num, language, or status values when the user did not provide them.

The Frevana endpoint schema currently exposes only q, page, num, language, and status; do not pass unsupported passthrough fields such as engine, api_key, output, no_cache, async, or zero_trace, and do not pass other unsupported Google Patents fields unless the Frevana endpoint schema is expanded first.

Response Shape

The API returns Frevana JSON. Common fields include:

• search_metadata
• search_parameters
• search_information
• organic_results
• summary
• pagination
• pagination

Output

• Success: the script validates that the response body is JSON, writes it to a file, prints the saved path to stderr, and prints the JSON to stdout
• Default file path: ./out/google-patents-search-<UTC timestamp>-<pid>.json
• With --output: the same JSON is written to the specified file path instead of the default path
• Failure: the script prints the response body or parsing error and exits non-zero

Notes

• Require --q or --query
• Do not call the script twice just to save and summarize results. It always saves the same JSON that it prints.
• Use --page only for page numbers 0 or greater
• Use --num only for result counts from 10 through 100
• Use --language when the user requests a patents language filter
• Use --status only for GRANT or APPLICATION
• If curl is missing, stop and tell the user to install curl
• If python3 is missing, stop and tell the user to install python3
• Do not echo the Bearer token back to the user
• Summarize patent title, publication number, assignee, inventor, priority date, filing date, publication date, grant date, snippet, link, and follow-up pagination when available unless the user asks for the full payload

Example Prompts

Chinese

• "Search Coffee in Google Patents"
• "Search Google Patents, q=(Coffee), status=GRANT, language=en"
• "Google Patents search AI coffee machine, page 0, 10 per page, and save the raw JSON"

English

• "Search Google Patents for (Coffee)"
• "Look up Google Patents for (Coffee) with status=GRANT and language=en"
• "Search Google Patents for AI coffee machine page 0 with 10 results and save the raw JSON"