Samson Laird

Loading projects…

• White-box research
  Gradient Untangler

  Local research harness that searches for the exact tokens that make an open-weight model start its answer the way you specify. White-box gradient attack on weights you already possess; remote APIs rejected.

  pip installable · local weights only

  Python PyTorch CLI
  Repository
• Agent forensics
  MASQ

  Local security CLI for AI agent stacks. Sits in the seat the model sits in: reads MCP catalogs and OpenAI-compatible HTTP the way the client does, reports what the model can actually see.

  cargo build --locked · bash scripts/demo.sh

  Rust MCP CI
  Repository
• OSCP toolkit
  Custom OSCP Tooling

  Read-only, OSCP-exam-safe tooling: cantina network recon orchestrator plus Star Wars-named enum, privesc-scan, hash-ID, and advisor scripts. Nothing modifies a target.

  python -m pytest tests/ -q · MIT

  Python CLI NSE bash
  Repository
• Reference
  LLM Injection Field Guide

  Catalog of prompt-injection and jailbreak techniques. Every attack card has a defense field.

  Live catalog · single index.html · CC BY 4.0

  HTML KaTeX zero-build
  Live Repository
• Defense library
  StegOFF

  Pre-ingest gate: scan text and files for steganography and prompt injection before they hit an LLM or agent. Clean what you can.

  python scripts/repro.py → REPRO_OK · offline pytest

  Python CLI API
  Repository
• Control plane
  Agentic DM Gateway

  Security control plane in front of LLM agents on private DMs. Who may talk, session unlock, kill switch, and message safety stay separate from the model.

  python scripts/repro.py → REPRO_OK · security unit tests

  Python Discord CI
  Repository
• Tripwires
  Agent Canary

  Tripwire detection for autonomous agents: file, MCP, and API honeypots that fire when something touches them.

  Public package + tests

  Python MCP honeypots
  Repository
• Eval lab
  Agent Trap Lab

  AI web-browsing agent trap lab: adversarial pages, StegOFF defense matrix, and an Ollama evaluation harness that publishes defended vs gap verdicts.

  Detector unit tests offline

  Python Ollama StegOFF
  Repository
• MCP auth
  nostr-mcp-auth

  Fail-closed NIP-98 authentication for HTTP MCP servers. No proof, no tools. Wrong key, bad sig, stale event, or body swap returns 401.

  nostr-mcp-auth quickstart · serve · call

  Python NIP-98 Nostr
  Repository
• Agentic research
  Equity Research Agent

  Natural-language equity research agent: plain English in, grounded market research out (price, technicals, fundamentals, SEC, news, X, charts, heatmaps, thesis memory).

  Public main branch · agent-focused default install

  Python Anthropic FastAPI yfinance Discord Docker
  Repository
• OSCP recon
  Cantina

  OSCP-legal network recon orchestrator: port scan, service enum plugins, multi-target timeouts. Enumeration only. No exploit auto-run.

  pytest suite

  Python pytest
  Repository
• Enum tooling
  vegadns

  High-concurrency subdomain enum and HTTP path discovery in one binary. Lab F1 measured against massdns / dnsx / ferox-class tools.

  docs/BENCHMARKS.md · lab suites only

  Rust DNS HTTP
  Repository
• Field notebook
  OSCP notes 2026

  OSCP field notebook, now the single vault for methodology that used to live in five separate repos: credential dumping, DLL attacks, UAC bypass, one-liners, and the bug bounty hunt ladder. Commands first.

  Public methodology vault · 512+ notes · consolidated

  Markdown Obsidian Python
  Repository
• Defense library
  Detection Defense Library

  One consolidated defense-engineering library: attack classes, the patterns that detect them, committed Sigma rules indexed to MITRE ATT&CK, and the Windows trust-boundary model in a single repo.

  Consolidated repo · patterns + sigma + boundaries + drivers + splunk

  Markdown Sigma Splunk SPL
  Repository
• Recon
  LM-Fingerprint

  Fingerprints the serving stack behind an OpenAI-compatible chat endpoint: tokenizer accounting, template offsets, JSON/SSE shape, role acceptance, named limits, gateway headers.

  pip install lm-fingerprint · MIT

  Python OpenAI API CLI
  Repository
• Agent skills
  Skill Library

  Own-authored agent skills, each a self-contained SKILL.md: testable procedure over vague ceremony. Original work only; no bundled-skill republishing.

  gitlab.com/WattoCyber/skill-library · MIT

  Markdown SKILL.md
  Repository

• Environmental Restoration LLC

  IT operations · 2022-present

  Two-person IT team, ~300 users nationwide, CMMC Level 2 environment. Own pieces of Device Control, BitLocker architecture, Autopilot modernization, and control-mapped policy work (NIST 800-171).

  Completed

• Offensive security

  OSCP in progress · lab and CTF

  OSCP in progress. 100+ machines rooted across HTB, Proving Grounds, TryHackMe, and VulnHub. Active Directory chains, Linux and Windows privilege escalation. Writeups published for retired boxes.

  In progress

• B.S. Cybersecurity

  Maryville University · 2025

  Three tracks: risk and compliance, offensive methods, and defensive operations. Graduated 2025.

  Completed