2.6k
GitHub Stars
3
Bundled Files
6 months ago
Catalog Refreshed
7 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill gemini-file- _meta.json282 B
- gemini_file_runner.py4.5 KB
- SKILL.md2.3 KB
Overview
This skill provides a Gemini CLI for one-shot Q&A, summaries, and content generation using positional prompts. It supports quick single-command queries and a file-output mode for large responses or when an explicit file is required. The tool includes model selection, output format control, and extension management.
How this skill works
Run the CLI with a prompt as a positional argument for one-shot answers (e.g., gemini "Your question..."). For large outputs or when you want a saved file, run the file runner script which writes results under an outputs/ directory and returns a JSON containing an absolute file_path. You can specify model, output format, file name, output directory, and whether to include content inline.
When to use it
- Ask a quick question or request a short summary directly from the terminal with a single command.
- Generate long articles, stories, reports, or processing-heavy outputs that exceed typical token limits.
- Parse long videos or dense transcripts and produce a structured file for downstream analysis.
- Automate batch generation or archival workflows that require a saved artifact for review.
- When you need JSON metadata plus a guaranteed absolute file_path for scripts or pipelines.
Best practices
- Use the positional prompt form for simple one-shot interactions to avoid interactive session overhead.
- Prefer the file-output mode for long content, video parsing, or when you need an auditable file.
- Default to gemini-3-flash-preview for large or complex content; test other models for shorter tasks.
- Include --include-content only when you need the response inline to avoid large stdout payloads.
- Run the CLI interactively once if authentication is required to complete the login flow.
Example use cases
- Quick Q&A: gemini "Summarize the key points of X article."
- Generate and save a long story: python3 .../gemini_file_runner.py --prompt "Write a 1000-word fiction" --output-file story.txt
- Video analysis: run the file runner with gemini-3-flash-preview to parse long transcripts into a structured summary file.
- Batch archival: script calls the runner for multiple prompts and collects file_path entries for each output.
- Format-controlled output: gemini --output-format json "Return a JSON object with fields a,b,c"
FAQ
The file runner always saves output under outputs/ and returns JSON that includes an absolute file_path. Inline content is optional with --include-content.
Which model should I use for long or dense inputs?
Start with gemini-3-flash-preview for long articles, dense transcripts, or video parsing, then experiment with smaller models for shorter tasks.
What if the CLI requires authentication?
Run gemini once interactively and complete the login flow; subsequent non-interactive runs will work with the saved credentials.