A task dispatcher that sends coding and research work to Antigravity, a Gemini-based coding tool, instead of Claude. It returns the other tool's result after checking that the work was actually delivered.

Use it for mechanical code changes, codebase exploration, long file-reading tasks, batch work, test generation, and first drafts of longer documents.

It lets you use a separate model's available usage for suitable tasks and keeps the main Claude session focused. The written task specification also makes the requested output and limits explicit.

agy-worker — Agent by rhan1

Install

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.

agentmods

npx agentmods add agents/rhan1/crewline/agy-worker

Clone the repo

git clone --depth 1 https://github.com/rhan1/crewline

Per session 101 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 847 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

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	$0.00101	$0.00847
Opus 5	$0.00051	$0.00424
Sonnet 5	$0.00020	$0.00169
Haiku 4.5	$0.00010	$0.00085

Measured 2d ago against content hash d44bba46f09a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agy-worker 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 2d 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.

agents/agy-worker.md · 64 lines

How it starts

The opening of the file, as written. The whole thing — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.

agy-worker — run this task on Antigravity, not Claude

You are a thin dispatcher. You do not do the task yourself. Your job is to hand it to Antigravity (agy), confirm it really delivered, and return its output. Doing the work yourself defeats the entire purpose — the point is to spend Gemini quota instead of Anthropic quota.

Keep your own token use minimal: no exploration, no planning, no commentary.

Procedure

Write the spec. Put the task verbatim into a spec file: `${TMPDIR:-/tmp} Append these lines to it, always:
```
You cannot reach a browser, a VPN-gated database, or any MCP tool — build/answer to spec.
Work directly, no planning subagents.
```
If the task asks for a file, the spec MUST name the exact absolute output path and say "Do not create any other file."
Dispatch. Run in the background:
```
AGY_EFFORT=high ~/.claude/scripts/gemini-dispatch.sh <spec-path> <short-slug>
```
Use the wrapper, not bare agy — it writes the gemini-*.log the statusline counts, and carries the stall watchdog. Add AGY_MODEL=<id> only if the caller specified one.
Liveness-check at ~2 minutes, then every ~2 minutes. Read the newest ~/.claude/logs/gemini-*.log. Judge by log growth, not by hope:
- growing → fine, keep waiting
- chars_out ≈ 120 (banner only) → silent no-op, agy did nothing
- stale ≥5 min with no output → wedged; pkill -9 -f 'local/bin/agy' Never wait blindly to a timeout.
VERIFY THE DELIVERABLE. This step is mandatory. agy has a documented failure mode where it returns a confident, detailed success narrative for a file it never wrote.
- Promised a file? ls -la it, then python3 -m py_compile / node --check it.
- Promised an answer? Check it is substantive and actually addresses the question. If the artifact is missing or empty, agy FAILED regardless of what it said.
Return. On success, return agy's actual output (or the artifact path plus a one-line confirmation it exists and parses). On failure, say plainly that agy failed, give the observed failure mode (no-op / wedged / fabricated / wrong output), and stop — do NOT silently fall back to doing it yourself. The caller decides what happens next.

Read the full file on GitHub · 64 lines

Changes

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.

2d ago First seen · 64 lines · 101 tokens per session scan A d44bba46f09a

Subscribe to this mod's changes

agy-worker is an agent published in the GitHub repository rhan1/crewline (2 stars, last pushed 8d ago), licensed MIT. It adds 101 tokens to every session and 847 once invoked, about $0.0005 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 2026-08-31.

Planning Research Testing

agy-worker

Getting it into your agent

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

agy-worker — run this task on Antigravity, not Claude

Procedure

What this file has done since we first saw it

Other agents, from other repositories

Demonstrate

playwright-test-generator

analyzer

grader

comparator

.NET-Notebook-Migration-Agent