Gauntlet — Evidence-Driven Adversarial Review

A deep-review protocol that attacks code, APIs, infra and tests for falsifiable defects, then traces and repairs their root cause.

Dev & CodingAdvanced★ 430⑂ 46AI score 8/10Last updated: Oct 8, 2026

What it does

Runs a disciplined loop: map → attack → substantiate → trace cause → repair → verify → re-attack.
Picks a risk tier (focused / standard / critical) with hard round budgets and explicit "clean re-attack" exit gates.
Demands evidence — an executable counterexample, source trace or authoritative reference — instead of a quota of stylistic nitpicks; unsupported suspicions are dismissed.
For confirmed defects it traces symptom → mechanism → enabling condition → missing invariant and hunts sibling bugs sharing that mechanism.
Bans green-washing: no weakened assertions, swallowed exceptions, arbitrary sleeps or disabled checks.
Ends with an honest status: REVIEW_COMPLETE, PASS_WITHIN_SCOPE, PASS_WITH_EXCEPTIONS, BLOCKED or BUDGET_EXHAUSTED — never "bug-free".

Who it's for

Developers and tech leads doing production-readiness or red-team reviews before shipping.
Anyone chasing a recurring bug that keeps coming back after shallow patches.
Teams touching auth, money, destructive data operations or cross-tenant boundaries.
People who want to stop an agent from declaring victory without proof.

Examples

Run a critical-tier, 6-round review on a refund API PR: reproduce a double-refund race, fix the missing idempotency invariant, add a regression test, re-attack from a new angle.
Read-only security pass on a multi-tenant service using only the trust-boundary lens, returning substantiated findings with evidence locations and unresolved hypotheses.
Standard-tier review of a flaky queue retry path that replaces "add a sleep" with a real fix to the timing invariant plus a property check.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/aiskillstore/marketplace/HEAD/pending/aakashh242/gauntlet/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)

Download the ZIP with the button below.
In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.

↓ Download ZIP

Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the pending/aakashh242/gauntlet folder from the GitHub repo aiskillstore/marketplace into my ~/.claude/skills/gauntlet/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/aiskillstore/marketplace.git /tmp/marketplace && mkdir -p ~/.claude/skills && cp -r /tmp/marketplace/pending/aakashh242/gauntlet ~/.claude/skills/gauntlet

⚠ This is a third-party skill. Check the source repository before installing.

Open a terminal.
Clone the repo to a temp folder: git clone https://github.com/aiskillstore/marketplace.git /tmp/marketplace
Create the skills folder: mkdir -p ~/.claude/skills
Copy the skill: cp -r /tmp/marketplace/pending/aakashh242/gauntlet ~/.claude/skills/gauntlet
Verify with ls ~/.claude/skills/gauntlet — you should see SKILL.md plus the references/ and templates/ folders (the skill relies on them).
Restart Claude Code, then ask something like: "Use gauntlet to do a deep adversarial review of this PR."
Optional: the local tracker needs Python 3.10+ with sqlite3 (python3 --version).
Say "review and fix" if you want bounded repairs; otherwise the skill stays read-only and just reports findings.

View source on GitHub ↗