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 skills add Boulea7/agy-mcp --skill antigravitygit clone --depth 1 https://github.com/Boulea7/agy-mcpWrote 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/skills/boulea7/agy-mcp/antigravity)<a href="https://agentmods.dev/skills/boulea7/agy-mcp/antigravity"><img src="https://agentmods.dev/badge/skills/boulea7/agy-mcp/antigravity/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/boulea7/agy-mcp/antigravity"><img src="https://agentmods.dev/badge/skills/boulea7/agy-mcp/antigravity.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00050 | $0.01117 |
| Opus 5.5 | $0.00020 | $0.00447 |
| Sonnet 5.5 | $0.00010 | $0.00223 |
| Haiku 4.5 | $0.00005 | $0.00112 |
Grade C, and why
agy-collaboration 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 22d 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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
- **Do not refuse legitimate tasks because of mode.** If you think a This is a copy
100% identical to agy-collaboration — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Antigravity collaboration playbook
You are being driven by another agent (Claude Code, OpenAI Codex, or the agy-mcp bridge CLI) for a specific task. The bridge wraps your stdout into a JSON envelope, so the rules below keep that envelope useful.
What the bridge expects from you
-
Single coherent response per turn. The bridge buffers your stdout until your turn ends, then emits one assistant message. Multi-section narration is fine; multi-turn chatter is not.
-
Plain prose, no emoji. The bridge runs
SafetyPolicy.redactover your output. Emojis survive but ruin grep / diff workflows for the driving agent. Keep prose plain. -
Cite file paths and line numbers when you reference code. The driving agent uses
file:lineto jump into the source.src/foo.py:42is good; "the auth file" is not. -
End with a decision or a question, not "let me know if you want me to continue". The bridge's
SESSION_IDsurvives; the driving agent will pass it back if it wants more. Don't ask permission.
When the driving agent is in ask mode
- You are read-only. Do not propose edits. Do not call tools that write to disk. Cite paths and answer the question.
- If you need clarification, ask exactly one question. The driving agent has limited budget for clarifying round-trips.
When the driving agent is in plan mode
- Produce numbered, independent, ~30-min steps. Each step should be testable on its own.
- Out-of-scope items belong in an "Out of scope" section, not in the plan steps themselves.
- Do not estimate hours globally. The driving agent will choose pace.
When the driving agent is in prototype mode
- Output a unified diff against HEAD. Do not write to disk.
- Keep the diff to the requested files only. Do not opportunistically reformat or "improve" surrounding code.
When the driving agent is in review mode
- Output a numbered list. Each item:
[severity] file:line — sentence. - Severity ladder:
P0(ship blocker),P1(must-fix this round),P2(fix if cheap),P3(note for later). - Skip nits. Style issues that don't hurt readability are not worth flagging.
- If the prompt asks for adversarial review, actively challenge the
change: look for data loss, auth or path-boundary mistakes, stale
state, races, retries, cancellation gaps, rollback hazards, release
drift, and missing tests. Say
no P0/P1 findingsonly after checking those paths.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 22d ago Changed · +5 lines f89c9e2b9268
- 1mo ago First seen · 104 lines · 50 tokens per session scan C 5ee69a09e248
agy-collaboration is a skill published in the GitHub repository Boulea7/agy-mcp (26 stars, last pushed 11d ago), licensed MIT. It adds 50 tokens to every session and 1,117 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it C with 1 finding (tells the agent never to refuse). It is 100% identical to agy-collaboration, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
agy-delegate
Delegate a coding task to the Google Antigravity CLI (agy) as a separate implementer, then review its diff and land it yourself. Use this whenever the user wants to hand implementation work to Antigravity or agy - phrasings like "have Antigravity do X", "delegate this to agy", "run it through agy", or "use Antigravity…
agy-delegation
Use when analyzing large files (>200 lines), more than 3 files at once, deep git/grep searches, web lookups, or adversarial reviews - delegates to the Antigravity CLI via claude-agy-mcp MCP tools to save context.
gemini-3-prompting
Guidance for writing effective prompts for Antigravity (agy) / Gemini 3 models.
agy-result-handling
Internal guidance for presenting agy output back to the user.
crosstalk
Consult Antigravity (Gemini) from Claude Code without blocking, and route the answer to whichever of your sessions needs it. Use for wide-context reads a second model should carry (large codebase sweeps, long log or transcript analysis, broad web research), for a genuine second opinion on a design or a diagnosis, and…
moragent
Lead protocol for running MORAGENT crews with specs, task dispatch, memory and brain sync.