agy-review
OctavianTocan/agy-review
Red-team automático para planos, código, escrita e decisões via CLI agy
Ver no GitHubSkill que integra o CLI Antigravity para fazer revisões adversariais de qualquer artefato — planos, designs, código, escrita, decisões, configs, ideias. Fornece crítica independente e rápida identificando suposições questionáveis, riscos e lacunas antes do commit.
Instalação
npx convez-ai OctavianTocan/agy-review1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Revisar designs de arquitetura antes de implementação
- Questionar decisões técnicas e de produto antes do commit
- Validar código, configs ou prompts em busca de falhas e riscos
- Critiquar escritas, argumentos e pitches para audiências técnicas
Veredito

Ferramenta prática e bem implementada para injetar pensamento crítico no fluxo de trabalho. Usa Gemini 3.5 Flash em modo thinking, roda em sandbox (não toca seu repo), e funciona com qualquer tipo de conteúdo textual via piping ou arquivo. Ideal para hardening antes de ship.
README
A tradução para o português deste README ainda está na fila.
agy-review
Adversarially review anything with the
agy(Antigravity) CLI before you commit to it — a fast, independent second opinion that pokes holes in whatever you give it.
What It Does
agy-review is an AI-agent skill that pipes content into the agy CLI's non-interactive print mode and returns a concise red-team critique: questionable assumptions, flaws, risks, gaps, failure modes, and a SHIP / REVISE / RETHINK verdict. It's meant to harden a thing before you commit to it.
It reviews whatever you give it — a plan or design is just one case. Also: a code change, an essay or doc, a product/architecture decision, an argument or pitch, a config, a prompt, a name, an idea. Raw code, numbered specs, and prose all work.
Defaults to Gemini 3.5 Flash in its highest thinking mode. The script runs agy from a throwaway empty directory, so it never reads or touches your repo — it just asks a question and prints the answer.
Skill Architecture
agy-review/
├── SKILL.md # routing + how-to (what it reviews, arguments, reading the verdict)
└── scripts/
└── agy-review.sh # wraps `agy --print`: builds the critique prompt, runs it, prints the result
Usage
# pipe content in
printf '%s\n' "$THING" | scripts/agy-review.sh -
# from a file, with context about what it is / goals
scripts/agy-review.sh -f design.md -c "must ship this sprint; Postgres only"
scripts/agy-review.sh -f essay.md -c "blog post for a technical audience"
# pick a different model
scripts/agy-review.sh -m "Gemini 3.1 Pro (High)" -f decision.md
Options: -f/--file, -c/--context, -m/--model, -h/--help.
Env: AGY_REVIEW_MODEL, AGY_PRINT_TIMEOUT (2m), AGY_HARD_TIMEOUT (150s).
Exit codes: 0 critique printed · 1 usage error · 2 agy not found or returned nothing.
Implementation note
agy's -p / --print / --prompt flag takes the prompt as its value, so every other flag (--model, --print-timeout) must come before it, with -p "$PROMPT" last. Put -p first and it silently swallows the next flag as the prompt. The script handles this; keep the ordering if you edit it.
Requirements
- The
agy(Antigravity) CLI on yourPATH(agy --version,agy models).
Installation
Install with Vercel's Skills CLI (recommended):
npx skills add OctavianTocan/agy-review
Or copy the directory into your agent's skills folder:
cp -r agy-review ~/.claude/skills/ # Claude Code (personal)
cp -r agy-review ~/.agents/skills/ # all local agents
License
MIT