Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/sudoax0n/gemini-watch/gemini-mdgit clone --depth 1 https://github.com/sudoax0n/gemini-watchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sudoax0n/gemini-watch/gemini-md)<a href="https://agentmods.dev/instructions/sudoax0n/gemini-watch/gemini-md"><img src="https://agentmods.dev/badge/instructions/sudoax0n/gemini-watch/gemini-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00766 | $0.00766 |
| Opus 5 | $0.00383 | $0.00383 |
| Sonnet 5 | $0.00153 | $0.00153 |
| Haiku 4.5 | $0.00077 | $0.00077 |
Grade C, and why
gemini-watch GEMINI.md scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
The script prints the working directory at the end: `Work dir: <path>`. If the user is done with their queries and won't ask follow-ups, delete it using `Remove-Item -Recurse -Force "<path>"` (or `rm -rf "<path>"` if in How it starts
The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Gemini Watch Extension
This extension gives Gemini CLI "eyes" to watch video inputs (URLs or local paths) natively. It downloads videos with yt-dlp, extracts frames with ffmpeg, and transcribes audio via captions or Whisper. It then presents these frames and transcripts so you can analyze the video.
Step 0 — Setup preflight (runs every invocation)
Before running /gemini-watch, verify dependencies and API keys:
py scripts/setup.py --check
- Exit 0 (Success): Proceed to Step 1 silently (do not announce setup is complete).
- Exit 2 (Missing binaries like ffmpeg, ffprobe, yt-dlp): Instruct the user to run the installer:
py scripts/setup.py. - Exit 3 (Missing Whisper key): Run the installer to scaffold the dotenv file, then ask the user for a Groq API Key (preferred — console.groq.com/keys) or OpenAI API Key (platform.openai.com/api-keys) and write it into
~/.config/gemini-watch/.envasGROQ_API_KEY=...orOPENAI_API_KEY=.... - Exit 4 (Both missing): Run the installer, then ask the user for a key.
Step 1 — Parse Input
Separate the video source (URL or path) from the user's question.
Step 2 — Run the Watch Script
Invoke the watch script passing the source verbatim:
py scripts/watch.py "<source>"
Useful Flags:
--start T/--end T— Zoom into a section. AcceptsSS,MM:SS, orHH:MM:SS. Auto-fps will scale denser for focused sections.--max-frames N— Lower the cap for a tighter token budget (default 80, max 100).--resolution W— Frame width in pixels (default 512; increase to 1024 only if reading small on-screen text).--no-whisper— Disable Whisper fallback.--whisper groq|openai— Force a specific backend.
Step 3 — Read the Frames & Audio
The watch script outputs a list of JPEG frame paths and an optional audio file path.
Use the read_file tool to read all of these JPEGs and the audio file in parallel.
- Visuals: Reading JPEGs allows you to see the video content natively.
- Audio: If a transcript is missing but an
audio.mp3file is provided, read it natively to listen or transcribe the content yourself.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 48 lines · 766 tokens per session scan C 9e68c54bbed4
gemini-watch GEMINI.md is an instructions file published in the GitHub repository sudoax0n/gemini-watch (1 stars, last pushed 3mo ago), licensed MIT. It adds 766 tokens to every session, about $0.0038 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
gemini-faf-mcp GEMINI.md
Instructions for Wolfe-Jam/gemini-faf-mcp, covering gemini.md — gemini-faf-mcp, test & verify, where things live, stack and before changing things.
gemini-faf-mcp CLAUDE.md
Instructions for Wolfe-Jam/gemini-faf-mcp, covering claude.md — gemini-faf-mcp, what this is, stack and context.
gemini-faf-mcp AGENTS.md
Instructions for Wolfe-Jam/gemini-faf-mcp, covering agents.md — gemini-faf-mcp, project, instructions for ai agents, context and stack.
self-command GEMINI.md
Instructions for stevenAthompson/self-command, covering self command extension, tools, 1. selfcommand, 2. geminisleep and 3. watchlog.
meteocontrol-mcp GEMINI.md
Gemini CLI instructions for nielsvbrecht/meteocontrol-mcp, covering meteocontrol extension context, core mandates & communication style, operational guidelines, 1. performance analysis & metrics translation and 2. proactive troubleshooting.
run-long-command GEMINI.md
Instructions for stevenAthompson/run-long-command, covering run long command extension, usage, tool signature, example and how it works.