code-review-gemini-swarm-with-ntm
Made by JeffGemini code review swarm via NTM, driven by the Antigravity CLI (agy). Use when "gemini review swarm", "agy review swarm", "code review with gemini/agy", or multi-round code auditing with Gemini agents.
Works With
Quality and Utility come from canonical skill metadata. Community ratings appear separately as stars when subscriber reviews exist.
Subscribe to download
Public previews are available, but downloads, CLI sync, and premium content require an active subscription.
Sign In to Subscribe- Full skill content with deterministic versioning
- CLI sync + local search index
- Access to Jeffrey’s premium library
Save to your account
Saved skills live in your cloud list. Run jsm sync on each machine to pull them locally.
What's New with this Skill
2Updates, enhancements, and release notes for this skill.
Enhanced: code-review-gemini-swarm-with-ntm, the readiness wait works again
Current NTM rejects bare integers for durations, so the swarm's readiness wait died at spawn with a missing-unit error before a single review round ran. The wait now passes --timeout=120s and the review loop starts cleanly.
- ntm --robot-wait --condition=idle --timeout=120s: NTM now requires duration units; the bare 120 form errors with INVALID_FLAG
New skill: Code Review Gemini Swarm with NTM
Orchestrate a swarm of Gemini 3.1 Pro agents (5 by default) through repeated review rounds, each cycling through study, deep exploration, and cross-review. Claude Code checks the swarm every 3 minutes, feeds the next prompt, and retires any agent that hits usage limits.
- Spawns N Gemini 3.1 Pro agents with hard Flash-model rejection: agents that hit usage limits are retired, never downgraded
- 7-prompt lifecycle per agent per round: study, then 3x (explore + cross-review), reusing the loaded context through prompt caching
- 3-minute CronCreate monitoring loop advances idle agents, retires rate-limited ones, and critically reviews output quality
- Prompt bank with 4 core prompts and 7 variants, plus sequencing patterns for small, standard, and large codebases
- Operator decision tree and quality rubric from real sessions: know when agents are finding real bugs vs. narrating code
This skill runs Gemini 3.1 Pro as a code-review swarm. It spawns a configurable number of agents via NTM, walks each one through a 7-prompt sequence per round (study the codebase, then three passes of exploring files for bugs and cross-reviewing code written by the other agents), then kills and relaunches them for the next round. A CronCreate heartbeat checks every 3 minutes to advance idle agents, retire any that offer a Flash downgrade, and report findings. It includes the exact Flash fallback strings (collected from 10+ real sessions), prompt variants and sequencing patterns for different c