Zero-exposure browser privacy desk, CLI, and Model Context Protocol (MCP) server.

• The Problem: Developers paste credentials, DB strings, and logs into frontier LLMs. Once sent to frontier models or public issue trackers, a secret cannot be un-sent.
• The Solution: Operates 5 client-side engines: text credential scrubber (8 detector classes + regex), zero-exposure prompt roundtrip with .aiscrub.json reconstruction key, zero-width Unicode watermark remover, C2PA EXIF/metadata wiper, and canvas redactor.
• Tech Stack: React 19 · TypeScript · Web Crypto API · Vite · Tailwind CSS · MCP Server · Node CLI

Launch AIScrubber AIScrubber GitHub

Browser-local cryptographic credential, passphrase, and pattern generator.

• The Problem: Online password generators transmit generated secrets across networks or rely on biased pseudo-random implementations.
• The Solution: Uses browser crypto.getRandomValues() with rejection sampling to eliminate modulo bias. Provides 4 modes: Random Password, Memorable Passphrase, Numeric PIN, and Custom Pattern with live bit-entropy calculation.
• Tech Stack: React 19 · TypeScript · Web Crypto API · Vite · Tailwind CSS v4 · GSAP

Launch SafeGen SafeGen GitHub

Credit-weighted semester GPA engine & Target What-If degree planner.

• The Problem: University grading scales vary widely. Ad-filled online calculators fail on custom credits, multi-semester weighting, and forward-looking GPA target forecasting.
• The Solution: Preloaded with university presets (VTU CBCS, Anna University, Mumbai University, KTU, JNTU, US 4.0) plus a custom scale builder. Includes a Target CGPA What-If planner calculating exact required semester GPAs.
• Tech Stack: React 18 · TypeScript · Vite · Tailwind CSS · LocalStorage Persistence

Launch GradeForge GradeForge GitHub

Sub-3-second local runtime, service, and environment readiness auditor.

• The Problem: Developers clone a repository and hit cryptic runtime mismatches, missing .env variables, stopped database daemons, or port collisions.
• The Solution: Automatically inspects project manifests (.nvmrc, package.json, pyproject.toml, go.mod, docker-compose.yml, .env.example) and generates single-line copyable fixes.
• Tech Stack: TypeScript · Node.js · pnpm Workspaces · tsup · Vitest · MCP Protocol · VS Code Extension

EnvPreflight GitHub npm package

Deterministic AI agent skills index — 277 verified skills across 12 focused categories.

• The Problem: AI coding agents guess when they hit domain-specific tasks, and long tool descriptions get truncated silently by model contexts.
• The Solution: A machine-indexed library of 277 agent skills across 12 category repos, plus standalone skillary-agents multi-agent orchestrator. Every description stays under 200 chars with explicit sibling boundaries and zero-indicator security checks.
• Tech Stack: Agent Skills Spec · SQLite FTS5 Search · Python 3.12 · Claude Code · Codex · Cursor · Gemini CLI

Skillary GitHub skillary-agents

Everything your bug report needs, in one command. Sub-2-second local diagnostic collector.

• The Problem: Filing a useful bug report takes 10 minutes of manually gathering logs, git states, package versions, OS info, and env config — developers skip half of it, leaving issues unactionable.
• The Solution: A zero-config CLI that bundles complete runtime context, sanitizes sensitive secrets with zero network calls, and copies a single markdown report ready to paste into GitHub or AI agents.
• Tech Stack: TypeScript · Node.js · AST Sanitizer · Vitest · tsup · CLI

Crashpack GitHub npm package

guillaumemeyer/watermarks-remover • Collaborator

• What it solves: Neutralizes invisible zero-width Unicode tracking vectors (\u200B, \u200C, \u200D, \uFEFF, \u2060) embedded by frontier LLM outputs.
• Contribution: Hardened the CLI pipeline and designed the AST-level regex sanitizer, transforming experimental code into a production-grade inspection engine with push collaborator access.

PrithvijitBose/Knowledge-Agent • Contributor

• What it solves: Hybrid vector-graph retrieval architecture for autonomous local reasoning agents.
• Contribution: Engineered multi-layer vector graph query pipelines, local embedding index integration, and autonomous reasoning loops for sub-second retrieval without cloud dependencies.

Graphify-Labs/graphify • Contributor

• What it solves: Turns codebases into persistent, queryable knowledge graphs for autonomous AI coding agents.
• Contribution: Contributed to the core v8 architecture, optimizing god-node detection, multi-file dependency cluster algorithms, and graph query traversal speed.

JuliusBrussee/caveman • Contributor

• What it solves: High-throughput codebase analysis and AST scanning for developer repositories.
• Contribution: Engineered deep repository scan heuristics and dependency profiling optimizations to eliminate bottlenecks across large monorepos.