THE GAUNTLET

Ship Track — Step 3

I am inevitable — and so is quality when you run the Gauntlet. Five rounds. Thirty agents. Nine universes. Every domain your project touches gets reviewed by specialists who do not pull punches. This is review-only — no building, no fixing. Just the truth about whether your code is ready to ship.

Thanos

How slash commands work: All VoidForge commands starting with / run inside Claude Code, not your system terminal. Launch Claude Code with claude in your project directory, then type commands like /gauntlet at the > prompt.

WHAT IS THE GAUNTLET

The Gauntlet is VoidForge's comprehensive review protocol. It does not write code — it judges it. Thanos orchestrates 30+ agents across QA, UX, security, architecture, and code quality to find every issue before your users do.

In your project directory, launch Claude Code and run the Gauntlet:

claude
> /gauntlet

Run it after a /build or /campaign completes. The Gauntlet produces a scored report with findings ranked by severity. Fix the critical issues, run it again, and repeat until you pass clean.

THE 5 ROUNDS

Round 1 — Discovery. Broad sweep across the codebase. QA, UX, security, and architecture agents each run independent audits and surface initial findings.

Round 2 — First Strike. Deep-dive into the highest-risk areas identified in Discovery. Specialists drill into specific files, flows, and edge cases.

Round 3 — Second Strike. A fresh set of agents reviews the same areas with different perspectives. No single reviewer's blind spots survive two independent passes.

Round 4 — Crossfire. Agents from different domains challenge each other's findings. Security questions UX decisions. QA questions architecture choices. Conflicts are resolved by evidence, not rank.

Round 5 — Council. The senior agents convene, deduplicate findings, assign severity scores, and produce the final Gauntlet report.

GAUNTLET MODES

Evaluating an existing codebase before a rebuild? Use the assess flag:

/gauntlet --assess

Assessment mode runs Rounds 1-2 only (Discovery + First Strike) with no fix batches. Findings are grouped by root cause instead of domain — you see the three stubs that all trace to the same abandoned migration, not twelve separate findings across six domains. This powers the /assess command.

Need speed? Use the quick flag for a 3-round pass:

/gauntlet --quick

Preparing for launch? The reckoning flag runs a 5-wave parallel review focused on launch readiness:

/gauntlet --reckoning

The full agent roster is deployed by default. Use --light for standard agents only (no cross-domain spot-checks), or --solo for lead agent only with zero sub-agents.