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.
git clone --depth 1 https://github.com/MarcosNahuel/antigravity-plugin-ccWrote 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/commands/marcosnahuel/antigravity-plugin-cc/transcribe)<a href="https://agentmods.dev/commands/marcosnahuel/antigravity-plugin-cc/transcribe"><img src="https://agentmods.dev/badge/commands/marcosnahuel/antigravity-plugin-cc/transcribe.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 5.1 | $0.00071 | $0.00788 |
| Opus 5 | $0.00036 | $0.00394 |
| Sonnet 5 | $0.00014 | $0.00158 |
| Haiku 4.5 | $0.00007 | $0.00079 |
Grade A, and why
transcribe scanned grade A with 0 findings 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 8d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Offload audio/video understanding to agy (Gemini is natively multimodal in audio + video; Claude
Code is not). Give it a local audio/video file or a YouTube/remote URL; get a faithful transcript
and a summary back, written to a file (cheap on Claude's context).
Raw user request: $ARGUMENTS
Phase 0 — Resolve source + kind (ONE Bash call)
Parse $ARGUMENTS: the first token that is a URL (http:///https://) OR an existing file
(test -f) is the source; everything else is an optional focus. If nothing resolves, ask
once: "¿Qué audio/video/URL transcribo?" and stop.
python - "$SOURCE" <<'PYEOF'
import sys, os, re, datetime
src = sys.argv[1].strip().strip('"')
AUD = (".ogg",".mp3",".wav",".m4a",".aac",".flac",".opus",".aiff")
VID = (".mp4",".mov",".webm",".mkv",".avi",".m4v")
low = src.lower()
if low.startswith("http"):
kind = "url"; addir = ""
base = re.sub(r"[^a-z0-9]+","-", re.sub(r"^https?://(www\.)?","",low)).strip("-")[:50] or "url"
else:
kind = "video" if low.endswith(VID) else "audio"
addir = os.path.dirname(os.path.abspath(src))
base = re.sub(r"[^a-z0-9]+","-", os.path.splitext(os.path.basename(src))[0].lower()).strip("-")[:50] or "media"
out = os.path.join("docs","agy","transcripts", f"{base}.md")
os.makedirs(os.path.dirname(out), exist_ok=True)
print(f"KIND={kind}\nADD_DIR={addir}\nWRITE_FILE={os.path.abspath(out)}\nSOURCE={src}")
PYEOF
Phase 1 — Transcribe (ONE agy subagent)
Spawn ONE antigravity:agy-rescue subagent in MODE: transcribe:
MODE: transcribe
CWD: <absolute current working dir>
KIND: audio|video|url
SOURCE: <file path or URL>
ADD_DIR: <dir of the source file, or empty for a URL>
FOCUS: <focus text or empty>
WRITE_FILE: <WRITE_FILE>
USER_TEXT:
(empty)
Phase 2 — Present
Read ONLY <WRITE_FILE> and present: the summary first, then the transcript path. Don't re-process
the media yourself.
Notes
- Gemini handles common audio (ogg/opus/mp3/wav/m4a/flac) and video (mp4/mov/webm/…); YouTube and other public URLs work without downloading.
- Long files (>~30 min) can hit the timeout — split with
ffmpegfirst if needed (the subagent raises the timeout for video/URL automatically). - agy
--printwrites nothing to stdout outside a TTY (issue #76); the transcript is read from the file.
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.
- 8d ago First seen · 68 lines · 71 tokens per session scan A 1f91e93b8731
transcribe is a command published in the GitHub repository MarcosNahuel/antigravity-plugin-cc (27 stars, last pushed 23d ago), licensed MIT. It adds 71 tokens to every session and 788 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
gemini-imagegen
Generate or edit an image via a Gemini (agy) session — Gemini's generateimage tool does what Claude cannot.
media
Understand an audio / video / image file — Antigravity (agy/Gemini) transcribes and analyzes it, returning a timestamped digest while the full transcript goes to a file.
watch-video
Watch and analyze a video file or YouTube URL — extracts frames and audio for understanding.
deep
A command for researching a topic from seven perspectives, including product strategy, user experience, technology, practice, academic work, history, and online discussion. It combines the findings into a web report, Markdown document, and Notion page.
agy-debug
Send an error/stack trace plus the relevant file(s) to Antigravity (agy) for root-cause analysis.
agy-review
Send the current change set to Antigravity (agy) for a senior-reviewer code review.