mehmet turac

@mturac ·Istanbul ·www.linkedin.com/in/mehmetturac

308 repos
33 followers
47 following

TypeScript 35%
Python 33%
HTML 12%
JavaScript 9%
Rust 7%

12k contributions in the last year

94-day current streak·94-day longest streak

Jul 2025

202122232425262728293031

Aug 2025

12345678910111213141516171819202122232425262728293031

Sep 2025

123456789101112131415161718192021222324252627282930

Oct 2025

12345678910111213141516171819202122232425262728293031

Nov 2025

123456789101112131415161718192021222324252627282930

Dec 2025

12345678910111213141516171819202122232425262728293031

Jan 2026

12345678910111213141516171819202122232425262728293031

Feb 2026

12345678910111213141516171819202122232425262728

Mar 2026

12345678910111213141516171819202122232425262728293031

Apr 2026

123456789101112131415161718192021222324252627282930

May 2026

12345678910111213141516171819202122232425262728293031

Jun 2026

123456789101112131415161718192021222324252627282930

Jul 2026

12345678910111213141516171819202122

Mehmet Turaç · @mturac Product judgment · engineering execution · agentic systems under production pressure I build the bridge between product judgment, engineering execution, and agentic systems that have to…

Mehmet Turaç · @mturac

Product judgment · engineering execution · agentic systems under production pressure

I build the bridge between product judgment, engineering execution, and agentic systems that have to survive real production pressure.

Over 15+ years in software, I have led and advised product, engineering, and strategy teams through
platform bets, architecture risk, organizational complexity, and delivery pressure. My work is not only
about building systems; it is about deciding which systems are worth building, aligning people around
them, and carrying them into production.

I think like a product owner, operate like a strategist, and still stay close enough to the code to know where the system will break.
That is the lane I like most: unclear problem, real users, brittle architecture, impatient timeline, no room for theatre.

In 2026, while continuing my professional work, I made a deliberate turn back toward open source:
to help public engineering teams, contribute to systems I respect, and make my judgment visible in the open. I focus on
small, reviewable patches in serious repositories: reliability fixes, edge-case handling, type safety, UI correctness,
CI repair, and maintainer feedback loops.

I care about teams and agents that can take responsibility in real codebases, not just look impressive in a terminal.
Occasionally loud in Turkish.

---

Open-source posture

I like public work that is small enough to review and real enough to matter: a failing edge case, a clearer contract, a safer default, a cleaner UI path, a less fragile test. Specific repositories and PR surfaces are listed further down; the top of the profile is just the operating thesis.

Private labs are active too: OSS scouting/factory systems, research automation, and contribution-quality gates that make the public work sharper without turning the profile into a list of private links.

Selected OSS proof

The work is public and reviewable. Each row links to completed contribution work submitted to the repository; current review and merge state stays on the PR itself. These are contribution links, not employment or affiliation claims.

| Maintainer surface | Contribution work | Public proof |
|---|---|---|
| Microsoft | Playwright annotation parsing and matcher extension reliability; governance rule validation | Playwright #41292, Playwright #41367, Agent Governance Toolkit #3299 |
| Meta / Facebook | ASTRYX theme artwork loading on the themes page | ASTRYX #3911 |
| Redis | Dynamic authentication credential refresh in the Rueidis Go client | rueidis #1008 |
| Elastic | Elasticsearch ESQL metadata and YAML parsing; Kibana reporting/Fleet; Elastic Agent cleanup; elastic-package testing | Elasticsearch #149151, Elasticsearch #152924, full Elastic PR history |
| Strapi | Core sessions, API generation, database healthchecks, URL-safe fallbacks, i18n, GraphQL, and admin reliability | public PR history |
| ChromeDevTools | MCP memory-leak debugging guidance and screencast frame-rate control | MCP #2330, MCP #2312 |
| Upstash | Context7 Node 18 compatibility and safer GitHub authentication handling | Context7 #2762, Context7 #2920 |
| nexu-io | Open Design and Looper reliability across daemon, provider, preview, UI, and long-running workflow paths | public PR history |

Builder shelf

The public repo shelf has two personalities: practical utilities that solve annoying engineering problems, and agentic systems that make autonomous work less fragile.

| Featured line | Projects | Why it exists |
|---|---|---|
| Content freshness | content-pulse, contentpulse-strapi, contentpulse-contentful | Find stale dates, version drift, and neglected content before users do. |
| Agent governance | claude-roundtable, promptguard, skill-hunter | Make coding agents check contracts, reuse skills, and pass quality gates. |
| Memory and proof | recall-mcp, proofbox, eval-kit | Turn vague agent confidence into persistent memory and repeatable checks. |
| Protocol tooling | mcpx, mcp-guard, mcp-scaffold | Build, proxy, and harden MCP servers without pretending every tool call is safe. |

Open the full public lab shelf · utilities, plugins, guards, agents, and research toys

🔍 content-pulse — Semantic Decay & Freshness Engine for PayloadCMS v3. Detect stale dates and outdated version references before your readers do. 14/14 tests passing.
🔍 contentpulse-strapi — ContentPulse for Strapi v5. Lifecycle hooks + admin widget for content freshness analysis.
🔍 contentpulse-contentful — ContentPulse for Contentful. Sidebar app with Forma 36 design system integration.
👻 simulacra — Ghost personas review your code before it ships. Find edge cases, race conditions, and business logic bugs before production.
🧠 sparse-attention-viz — MiniMax-style Top-k sparse attention visualizer. CPU-only, no GPU required.
🪑 claude-roundtable — Multi-agent governance for Claude Code. Council deliberates, votes, dispatches, and enforces quality gates.
⏱️ leyla-scheduler — Durable, session-aware task scheduler in Rust. Jobs survive disconnects.
🔎 skill-hunter — Pre-execution layer that makes agents check for existing skills before building from scratch.
🛡️ promptguard — Audits prompts as behavioral contracts. Pre-write guard for agents that ship code.
🧠 recall-mcp — Shared brain for AI agents. SQLite-backed persistent memory over MCP.
👵 moooom-claude — 10 cultures of moms nagging your Claude Code to drink water and sit up straight.
🐍 hermes-supercode-skills — 13 production-grade Claude Code skill modules: database, auth, observability, infra, security, deployment, debugging, API design, and more. Also available for OpenAI Codex.
🧰 pluginpool — Ten focused Claude Code plugins for everyday developer productivity. Each one is its own MIT-licensed repo: commit-narrator, pr-storyteller, test-gap, deps-doctor, env-lint, secret-guard, standup-gen, todo-harvest, flaky-detector, changelog-forge. 89 hermetic tests · Python 3 stdlib only.
♿ a11y-ci — Automated accessibility audits for Claude Code. Static WCAG 2.1 checks on HTML, JSX/TSX, and Vue SFC — violations annotated in-source with line numbers and fix suggestions.
📐 adr-sync — Keeps Architecture Decision Records aligned with actual code changes. Detects drift between ADRs and implementation before it compounds.
💰 agent-budget — Track token spend and enforce cost budgets for LLM agent runs.
🔗 api-contract — Validates frontend API calls against backend schemas, detects breaking changes before deploy, generates typed TypeScript clients.
⚡ build-cache-analyzer — Analyzes build cache hit rates across CI runs, finds what's busting your cache, and optimizes build times.
🏺 code-archaeologist — Uncovers intent and evolution in legacy codebases through git history, comments, and structural patterns.
👁️ codewatch — Automated PR review and security auditing for Claude Code.
✅ config-schema-validator — Validates YAML, TOML, and JSON config files against schemas and suggests auto-fixes for common misconfigurations.
🗺️ contract-atlas — Maps the implicit contracts in your codebase. You find out about them when they break — this finds them first.
💸 cost-watch — Token spend tracking and budget alerts for Claude Code sessions.
🚫 dead-end-finder — Discovers unreachable code, unused exports, and orphaned API endpoints that accumulate silently.
📦 dep-drift — Detects when your lockfile diverges from package.json, flags silent transitive dependency updates, and catches license incompatibilities before they become compliance issues.
🕸️ dependency-graph-viz — Builds import dependency graphs, detects circular dependencies, and analyzes deep chains that create brittle architectures.
🚀 deploy — Deploy to Vercel, Railway, Fly.io, or Render directly from your editor.
🍴 devfork — Git branch for your entire running dev environment. Switch context without killing your server.
📄 docs-sync-checker — Detects documentation drift: when docs describe functions, CLI commands, file paths, or config keys that no longer exist or have changed.
🧪 eval-kit — Assertion-based prompt evaluation for LLMs. Test AI outputs like unit tests, run from the terminal.
🚩 flag-ops — Manage feature flags: scan for orphaned flags, track rollout status, and safely remove flags when done.
🎯 flaky-test-hunter — Detects, quarantines, and fixes flaky tests by analyzing CI run history and test timing patterns.
🌍 i18n-sync — Keeps translation keys in sync across locales. Detects missing, extra, duplicate, and untranslated keys. Generates type-safe stubs and prunes orphaned entries.
🚨 incident — Production incident response for Claude Code. Triage fast, diagnose deep, fix correctly, document thoroughly.
⚖️ license-guard — Enforces license compliance policy on dependencies, generates SBOM, and blocks incompatible licenses before they reach production.
🧹 llmfmt — Normalizes LLM output: extracts JSON, code blocks, strips markdown, removes thinking tags, and unwraps common prose prefixes.
🏗️ mcp-scaffold — Scaffold production-ready MCP servers in seconds — stdio, multi-tool, and HTTP templates included.
🧠 memex — Personal semantic memory for claude.ai. Save notes, links, code snippets, and ideas — search and recall them in any conversation via a knowledge graph.
🗄️ migrate-safe — Analyzes database migrations for destructive operations and safety issues before you run them.
🔍 perf-lens — Identifies N+1 query patterns, redundant React re-renders, and bundle bloat before they hit production.
🔬 promptlab — Zero-backend visual prompt engineering studio. Multi-model comparison, variable interpolation, eval assertions, and shareable URLs.
🔄 refactor-preview — Simulates refactoring impact with blast radius analysis before applying changes — see exactly what would break.
⏮️ replay — Records and replays Claude Code agent runs. Every bash command, file read/write, and tool call logged to SQLite. Inspect, diff, and audit any past session.
⚗️ schema-forge — Generates Zod schemas from JSON examples, OpenAPI specs, and JSON Schema definitions.
📝 semantic-release-notes — Auto-generates human-readable changelogs and release notes from conventional commits and PR titles.
🔭 test-scout — Finds untested code paths and generates targeted unit tests by analyzing coverage gaps and control flow.
🏷️ type-coverage — Tracks TypeScript type safety across your codebase. Surface where any is spreading.
🎨 open-design — Local-first open-source alternative to Anthropic's Claude Design. 19 skills · 71 brand-grade design systems · generates web, desktop, mobile prototypes, slides, images, and video. Runs on Claude Code, Codex, Cursor, Gemini, OpenCode, Qwen, Copilot, Hermes, and Kimi CLI.
🔬 proofbox — Verify AI-generated technical recipes before engineering teams trust them.
🌊 weave — Visual AI-agent flow builder with real LLM execution and Python export.
🔀 mcpx — MCP gateway that proxies and routes multiple MCP servers behind one endpoint.
🛂 mcp-guard — Policy firewall and security scanner for MCP servers and agent tools.
🧹 debtlint — Vibe debt scanner: AI attribution and code ownership tracker.
🎭 **[multichar-studio](h

…

All public repos (200)

Show forks Show archived

More creators on gitmyhub

gustavoguanabara yyx990803 claude gaearon peng-zhihui