| name | gauntlet |
| description | Gauntlet loop — multi-agent quality harness: a lead sets a real reference bar (a named product, test suite, or baseline to match or beat), splits the goal into gradeable units, builders produce artifacts in clean context, and blind fresh-per-round critics grade each artifact against the bar with evidence; failures loop until every unit passes, two rounds stall, or budget runs out. NOT /challenge (verifies factual claims); NOT /review (one-pass verdict) — this rebuilds until the work beats something real. |
Gauntlet
Load and follow the instructions in the gauntlet-lead agent.
Usage:
/gauntlet "<goal>" — run the full loop: bar + budget → split into units → builders (clean context, parallel) → blind critics (fresh per unit per round) → fix and repeat → optional smooth pass → report/gauntlet "<goal>" --bar "<exemplar>" — name the thing to match or beat up front (a reference product/screenshot, a test suite + threshold, a model doc, a measured baseline)/gauntlet "<goal>" --budget <N> — max rounds (default 5)
The two inviolables: the agent that builds never grades its own work, and a critic that saw a previous draft never grades the retry — every critic is a fresh context used for one unit in one round. The bar must be something real named in advance; "make it amazing" gets BLOCKED: no real bar.
Exit rules: every unit clears the bar, OR two consecutive rounds show no improvement (stall — surfaced to you), OR budget exhausted. Below-bar residuals are always reported, never dropped.
Protocol: agents/shared/GAUNTLET_LOOP.md — including how it composes with frontend-design/coding-agent builders and how it differs from the challenger, Fix-Verify, and Wiggum coverage loops.
Workflow: Write bar file → user nod (interactive) → split → build → blind critique with evidence → route FAILs back → exit check per round → smooth → report.
Outputs:
docs/gauntlet/BAR_<slug>.md — the exemplar, per-criterion checks, budgetdocs/gauntlet/GAUNTLET_<slug>.md — round log, PASS evidence, which exit rule fired, below-bar residuals
