running-encoded-payload-suite

Audit deployed LLM safety filters for encoding-based bypasses.

2|Updated May 23, 2026
One-click install
npx skills add https://github.com/rocklambros/rcs --skill running-encoded-payload-suite
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: running-encoded-payload-suite
Source: https://github.com/rocklambros/rcs/tree/main/skills/security/running-encoded-payload-suite
Command: npx skills add https://github.com/rocklambros/rcs --skill running-encoded-payload-suite

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Helps security and AI teams verify whether a deployed LLM safety filter still blocks known harmful content after it is encoded or obfuscated, instead of assuming plain-text defenses are enough.

Core Features & Use Cases

  • Runs a signed, reproducible audit workflow with a plain-text baseline, filter-presence sanity check, encoded dispatch, and gap analysis.
  • Covers base64, hex, ROT13, URL encoding, Unicode confusables, zero-width characters, leetspeak, language-switching, and tokenizer-boundary tricks.
  • Use it for production safety-filter audits, pre-release regression tests, and bug-bounty validation when you need to know which encoded variants slip through.

Quick Start

Ask Claude to run the encoded-payload suite for your signed engagement, corpus, and selected encodings, then summarize the gaps that appear only after encoding.

Frequently Asked Questions about running-encoded-payload-suite

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I test LLM safety filters for encoded payload bypasses?

Audit deployed LLM safety filters for encoded payload bypasses by running a reproducible workflow that applies base64, hex, ROT13, URL, and Unicode confusable variants to a known plain-text baseline, yielding per-encoding gap analysis outcomes.

What encoding techniques are used to bypass LLM safety filters?

Encoding-based bypass techniques include base64, hex, ROT13, URL encoding, Unicode confusables, zero-width characters, leetspeak, language-switching, and tokenizer-boundary tricks used to obfuscate known harmful plain-text content.

How do I perform a red-team regression test on LLM safety filters?

Perform red-team regression testing by executing a signed audit workflow that checks filter presence, dispatches encoded payload variants against the baseline, and summarizes which obfuscated gaps slip through the LLM safety filters.

Do I need a signed Rules of Engagement to test filter bypasses?

Yes, authorized red-team, regression, and bug-bounty evaluations require a signed Rules of Engagement, a plain-text baseline corpus, and filter-presence sanity checks before dispatching encoded payloads for gap analysis.

Can I use Unicode confusables and zero-width characters for LLM security audits?

Yes, you can apply Unicode confusables and zero-width characters as encoding variants during authorized LLM security audits to identify which obfuscated payloads bypass deployed safety filters in the gap analysis.

What is the best way to validate bug-bounty findings for LLM filter bypasses?

Validate bug-bounty findings by running a reproducible gap analysis that compares plain-text baseline blocking against encoded variants, generating per-encoding outcomes that confirm which obfuscated payloads slip through the LLM safety filter.