Interactive ML project scaffolding. Supports single-model and multi-model projects with Rich terminal beautification, dependency-aware pipeline orchestration, complete Python package generation, and 5 standard CLI commands.
Interactive onboarding wizard that integrates worktree-wizard into existing projects — analyzes codebase, generates docker-compose with wt.* labels, sets up hot-reload, volume mounts, and justfile integration.
Déploie ta branche sur playground (/preview), remet playground à zéro = copie de develop (/clean), ou pousse tes données locales sur playground (/push-data). Zéro connaissance git ni Postgres.
Expert guidance for FoundryVTT v14 system and module development — data models, Active Effects V2, ApplicationV2 sheets, Regions, Scene Levels, hooks, testing, and i18n. Includes a proactive code reviewer for deprecated patterns and common mistakes, plus a V13-to-V14 migration reference. Not affiliated with Foundry Gaming LLC.
Real-time ESLint diagnostics for Claude Code via a lightweight LSP server. No vscode-eslint dependency. Supports ESLint v9+ flat config. Bundled with @eslint/mcp for AI-callable autofix.
Brutal but brilliant code reviewer with the personality of a celebrity chef. 8 modes, 6 personalities, auto-fix, shareable roast cards. Real findings with real jokes.
Debug and inspect GetStream feed subscriptions and activities from Claude — check who follows what, inspect activity fan-out, and repair missing notifications.
GitHub CLI MCP servers wrapping gh for pull requests, issues, CI runs, jobs, commits, search, labels, projects, and reviews. Read server (always active) with 31 tools and write server (opt-in) with 25 tools. Includes SessionStart hook that injects MCP tool directives and PreToolUse hooks that enforce MCP tool usage. Configuration-optional: works without config when gh is authenticated.
Generator-agnostic verifier for designed metal-binding proteins — geometry + bond-valence + CSD + physics/MLIP + co-fold consensus, exposed as an MCP tool and a /verify-binder command. Triage designs to wet-lab; score them as an RLVR reward.
Run intent-verify check or map from Claude Code for an explicit spec and repository. Results are advisory lexical evidence, never acceptance authority.
Git-native notes and tasks layer for AI agents — durable, synced work items, design decisions, handoff docs, journals, runbooks, investigations, approved plans, and a sprint/project grouping layer stored as objects in the git ODB on refs/cc-notes/*.
Opt-in Claude Code hooks: refuse attribution lines in commits and PRs, refuse staging COMMITS.md, refuse commits and pushes that carry a secret, refuse issue and PR bodies past their length budget, warn on commits in a main checkout, refuse a subagent call that would silently inherit the session's model, restart idle sessions onto an updated Claude Code, and route work to cheaper models while the usage allowance burns too fast.
Share localhost dev servers at public HTTPS URLs. Adds the localcloud MCP server (share_port, get_url, list_shared, stop_sharing, auto_share), a skill that teaches Claude when to share, and a SessionStart hook that auto-shares dev servers in cloud sessions.