antiRea
SDK antiRea 3.0.0: evidence-led, server-side TypeScript toolkit that meters dataset access per identity and detects coordinated bot abuse - 1056 automated tests (1054 passing, 0 failing, 2 skipped) on Linux x64 with Node 24.22.1; release gate 17/17.
Free
Works with the AI Agents you already use
antiRea
Example session with this skill installed
Wire up a detector to catch bot fleets that join our live rooms on a fixed timer and alert us if a coordinated herd is confirmed.
- Read your context and instructions
- Compiled the antirea
I have integrated the createHerdPhaseSignal module to detect coordinated bot fleets driven by shared schedulers. The implementation includes an evaluation loop that checks the cohort every 5 minutes against the lab-validated 24-hour window.
Herd Detection Integration
import { createHerdPhaseSignal, HERD_PHASE_ROOM_COHORT_WINDOW_MS } from "@dannyphan190/anti-rea";
// Initialize the signal for the specific live room
const roomJoinHerdSignal = createHerdPhaseSignal();
/**
* Hook this into your room join event handler.
*/
export function onUserJoin(accountId: string) {
roomJoinHerdSignal.observe({
identity: accountId,
ts: Date.now()
});
}
/**
* Background task to evaluate the fleet status.
* Results are advisory (enforcement: "none").
*/
export async function checkRoomHealth() {
const result = roomJoinHerdSignal.evaluate({
cohortWindowMs: HERD_PHASE_ROOM_COHORT_WINDOW_MS
});
if (result.confirmed) {
console.warn(`[Anti-Rea] Coordinated herd confirmed: ${result.memberCount} members.`);
// The host application decides the response.
// Recommended: apply createVerificationGrace to NEW join attempts.
return {
status: 'alert',
fleetMembers: result.members, // Named because cohort is confirmed
confidence: result.score
};
}
return { status: 'clear' };
}
Next steps
- Install the dependency using
npm install @dannyphan190/anti-rea. - Connect the
onUserJoinfunction to your WebSocket or REST join endpoint. - Implement a
createVerificationGracehandler to step-up new users whencheckRoomHealthreturns an alert.
Connects securely to your tools. The creator never sees your data.
About this skill
A normal web app is wide open to bulk extraction: in antiRea's tests, one account and a single export call walked away with an entire 10,000-record dataset in under 200 requests, quietly. With antiRea wired in, data is metered per record - one identity can only ever receive a capped slice (20% of the dataset in the test setup), so full extraction stops being an export call and becomes an identity problem. In the measured v2.5.0 classic round, the REA-style run that took one account unprotected was stopped at 5% of the dataset, and a scripted slicing crew that insisted on finishing needed 25 identities and about 19 minutes, flagged by the herd detector and the evidence engine on the way.
antiRea combines per-identity record budgets, priced exports and session-chained cursors, RFC 9421 Web Bot Auth, and a small herd detector so that many light, plausible accounts are read together instead of one by one. Its evidence engine then scores each account against every law - for example, export velocity: is a single account exporting at an unhuman pace? - and applies a step-up that fits the risk, first asking the user to prove they are human, and naming nobody until a cohort confirms. Release 3.0.0, the farm-calibration release, adds tier-2 conversion: a persistent herd flag with independent corroboration steps an identity's record budget down toward the 500-record floor - never below it, never a hard block - while a first decoy touch means taint plus step-up, all under a fleet-wide false-positive budget. It also adds bulk capability keys (the head of a dataset stays open, the tail needs a server-issued key, and the anonymous pool is capped at 6 keys per 24h across the fleet), an opt-in pin statistic (FC-07), a queue-stuffing axis for the vote-brigading law, and a self-disarm fix for room-hop and reconnect-storm (re-measured 20/20 on time, false positives unchanged). SDK antiRea 3.0.0 ships with 1056 automated tests (1054 passing, 0 failing, 2 skipped) on Linux x64 with Node 24.22.1; release gate 17/17. Everything is advisory: the skill reports a confidence and a recommended step-up; your app decides whether to challenge, throttle or ignore, so every deployment stays compliant and auditable.
Known limitations: (1) this SDK protects server-side data access, not browser rendering, UI interactions or client-side scraping; (2) rate limits and identity attribution depend on the host app correctly identifying callers - shared accounts, NAT and proxies can blur attribution; (3) the evidence engine requires meaningful traffic volume before cohort-level signals become reliable; (4) herd detection flags coordinated patterns, not individual intent, and cannot distinguish a legitimate coordinated campaign (such as a product launch) from a bot fleet without host context; (5) no automated tests target the Web Bot Auth signature endpoint yet.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
All signals are advisory: antiRea measures and reports, the host application decides. It cannot prevent anyone from inspecting code delivered to their browser, and it does not stop a determined, well-resourced rebuilder.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Free with an Agensi account
- Software
Node.js