Poorvith M P poorvith-mp

POORVITH M P

I Foresight before small problems become expensive.

Tools Architect • AI Agent Systems • Local-First & Privacy Engineering
Hassan, Karnataka, India

🏛️ About Me & The Foresight Principle

I notice the friction people learn to tolerate, turn it into a focused tool, and keep the engineering decisions visible from first check to working proof.

Most software failures don't happen because of complex math — they happen because small warning signs were ignored: a credential pasted into an LLM prompt, a missing environment variable crashing a dev container, or semester grade calculations lost in rigid spreadsheets.

I build browser-local utilities, developer tools, and modular AI agent systems that solve these exact points of friction before they turn into costly emergencies.

🛡️ Foresight First
_{Catching leaks, conflicts, and decay before staging or production}

🔒 Zero-Exposure Privacy
_{100% browser-local cryptography and offline-capable CLIs}

⚡ Useful Before Impressive
_{Sub-3-second runtimes, zero-config setups, and honest limits}

🔍 Visible Trade-Offs
_{Open documentation, verified benchmarks, and zero telemetry traps}

🚀 My Contributions

AIScrubber

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

View architectural highlights

100% in-browser Web Crypto execution with zero network telemetry
Claude & LLM invisible watermark remover (\u200B, \u200C, \u200D, \uFEFF, \u2060)
Terminal batch execution via npx aiscrubber
Native MCP server integration for Claude Desktop, Cursor, and Codex

SafeGen

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

View architectural highlights

Cryptographic randomness via Web Crypto API with zero server calls
Live bit-entropy scoring and offline brute-force crack time models
Local-only searchable copy desk up to 50 items with 1-click purge
Fully responsive tactile interface with GSAP micro-interactions

GradeForge

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

View architectural highlights

Multi-semester credit-weighted SGPA and cumulative CGPA calculation
Target CGPA 'What-If' planner for future semester planning
CSV data export, JSON backup, and clean printable transcript renderer
100% offline browser storage with zero account signups

EnvPreflight

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

View architectural highlights

Multi-runtime validation for Node.js, Python, Go, and Rust
Background daemon checks: Postgres (5432), Redis (6379), MySQL (3306), MongoDB (27017), Docker
Strict zero-leakage .env variable key auditor (secret values are never read or logged)
Model Context Protocol (MCP) server for Claude Code and Cursor agents

Skillary

The Universal AI Agent Skills Ecosystem — 315 modular agent instructions.

The Problem: AI coding agents lack structured, deterministic domain knowledge, resulting in shallow answers and repetitive manual prompting across projects.
The Solution: A standardized repository index of 315 agent skills across 12 domain packages compatible with Claude Code, OpenAI Codex, Cursor, and Gemini CLI. Adheres strictly to the Agent Skills Specification.
Tech Stack: Markdown Architecture · Agent Skills Spec · Progressive Execution Index · pnpm Workspaces

View architectural highlights

12 Categorized Domain Packages: Developer, Marketing, Design, Business, Finance, Writing, and more
Direct terminal installation via npx skills add poorvith-mp/skills-[category]
Built-in multi-sweep anti-AI tell standards and strict execution checklists
Seamless interoperability across leading agent frameworks

Crashpack

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

View architectural highlights

Zero network calls: completely offline execution with zero telemetry
Mandatory redaction pipeline: auto-strips AWS keys, Stripe tokens, private keys, and passwords
One-click GitHub Issue pre-fill (npx crashpack --issue)
Command wrapper mode (npx crashpack --wrap "command") capturing failed process buffers

🌐 Groundbreaking Open Source Contributions

Upstream core contributions, security hardening, and knowledge graph systems built for the ecosystem.

🔒 watermarks-remover

guillaumemeyer/watermarks-remover •

The Breakthrough: 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.

🧠 Knowledge-Agent

PrithvijitBose/Knowledge-Agent •

The Breakthrough: 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

Graphify-Labs/graphify •

The Breakthrough: 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.

⚡ caveman

JuliusBrussee/caveman •

The Breakthrough: 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.