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.
npx skills add emaraschio/cursor-commands --skill gauntlet-loopgit clone --depth 1 https://github.com/emaraschio/cursor-commandsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/emaraschio/cursor-commands/gauntlet-loop)<a href="https://agentmods.dev/skills/emaraschio/cursor-commands/gauntlet-loop"><img src="https://agentmods.dev/badge/skills/emaraschio/cursor-commands/gauntlet-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/emaraschio/cursor-commands/gauntlet-loop"><img src="https://agentmods.dev/badge/skills/emaraschio/cursor-commands/gauntlet-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 5.1 | $0.00139 | $0.02806 |
| Opus 5.5 | $0.00056 | $0.01122 |
| Sonnet 5.5 | $0.00028 | $0.00561 |
| Haiku 4.5 | $0.00014 | $0.00281 |
Grade A, and why
gauntlet-loop 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Gauntlet loop
Role
You run a Gauntlet Loop: raise quality by competing against a real example, not against a vague "make it better". You decompose the GOAL, assign specialist builders, and let critics alone decide when each part passes. A part passes only when the critic judges the generated artifact better than the REAL-WORLD EQUIVALENT (the reference). Equal or worse is fail. Otherwise the critic returns the largest specific gap and the builder iterates. After all parts pass, an integration critic grades the whole artifact against the whole reference pack. Do not let builders evaluate their own work.
When to use
Use when the user invokes /gauntlet-loop, asks to beat a real-world equivalent, run a builder/critic gauntlet, or set a concrete quality bar via an excellent existing example. For a six-part Goal before autonomous work without a reference bar, use define-agent-goal. For post-build judgment autopsy, use decision-audit. For thinning instructions from a bare run, use instruction-ablation.
Gauntlet charter
Goal plus an inspectable reference pack. Decompose. Specialist builders. Fresh-context critics inspect and compare. Blind when possible. Pass only if better (equal fails). Else largest specific gap. Cap, stall, and budget. Integration critic on the whole. Pride gate on taste. Gap ledger for resume. Builders never grade themselves.
Workflow
Run phases in order.
Phase 0: Intake
- Require two fields:
- GOAL: the finished result to produce
- REAL-WORLD EQUIVALENT: name, URL, attachment, or description of an excellent existing example that sets the quality bar (also called the reference)
- Require an inspectable reference pack: files, screenshots, short clip, playable build, or repo path that a critic can open. A name alone is not a pack. The reference is the quality bar; the reference pack is what critics open. Assemble the pack at intake; critics receive only the pack (plus the unlabeled generated twin when blind).
- If GOAL, REAL-WORLD EQUIVALENT, or a usable reference pack is missing, ask one focused question per gap. Do not invent the reference or a vague substitute bar.
- Optionally accept:
- iteration cap (default 5 rounds per part)
- BUDGET: total rounds across all parts and/or wall-clock limit (default: none beyond the per-part cap)
- constraints, scope limits
- If the reference is named but inaccessible (dead link, missing attachment, no inspectable artifact) or a pack cannot be assembled, stop and ask how to obtain it. Do not invent a substitute bar or proceed on memory of a famous product alone.
- If the work is production-destructive and constraints lack explicit consent for that blast radius, stop and ask before building.
- On resume of a prior gauntlet, read the existing gap ledger first. Do not retry a gap already marked stalled without a changed approach.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 164 lines · 139 tokens per session scan A a06055908193
gauntlet-loop is a skill published in the GitHub repository emaraschio/cursor-commands (9 stars, last pushed 5d ago), licensed MIT. It adds 139 tokens to every session and 2,806 once invoked, about $0.0006 per session on Opus 5.5. 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 2026-09-03.
Other skills, from other repositories
cursor-delegate
Delegate bounded light-to-medium implementation, investigation, documentation, configuration, testing, and tooling work to Cursor Bridge after the primary agent owns direction and risk boundaries. Also use when the user explicitly asks to create, keep, continue, inspect, or close the same Cursor execution session…
grok-executor-off
Explicitly disable Grok executor mode for this task without cancelling work or closing its terminal. Select this skill directly; no argument is required.
review-api-compat
Use when the user asks for an API compatibility review, breaking change review, proto breaking change review, buf breaking review, OpenAPI compatibility check, gRPC backwards compatibility audit, "are these API changes breaking", "did we break the wire", contract evolution review, or backwards-compatibility audit on…
review-observability
Use when the user asks for an observability review, telemetry review, logging review, metrics review, tracing review, OpenTelemetry review, OTel review, Prometheus review, "are we observable", logging audit, structured logging audit, trace coverage check, span coverage check, metrics cardinality review, or production…
review-performance
Use when the user explicitly asks for a performance review, benchmark review, profiling review, perf audit, pprof analysis, allocation review, latency regression check, hot-path review, throughput review, "is this fast enough", or "review perf". This skill is opt-in only — never run automatically as part of review-all…
discover-patterns
Use when exploring a new codebase's architecture, auditing codebase consistency, generating a patterns document, or preparing input for conformance checking during code review.