mbadali25

by Matthew Badali

A small, honest virtual dev team: context-isolated agents, deterministic gates, file-backed work state. Optional Codex QA, optional Jira or ServiceDesk Plus, optional Obsidian Kanban tickets, optional Obsidian memory.

by Matt

Gizmoduck — run Nuclei vulnerability scans on websites and hosts, diff against previous scans, and turn findings into triaged reports and ServiceDesk Plus tickets. Runs on WSL/Linux and Windows.

by Matthew Badali

Local models on your own GPU via Ollama, in two halves: three MCP tools that index a repo or folder into an on-disk vector store, search it, and answer from it inside your current session; and a `localgpu` shell/proxy that launches a separate Claude Code session running entirely on a local chat model, through a proxy translating the Anthropic Messages API into Ollama's chat API. Six commands cover setup, health checks, indexing, search, ask, and an honest account of which crew roles a local 7B can and cannot take over. No hooks and no agents; nothing leaves 127.0.0.1.

by Matthew Badali

Make one or more Obsidian vaults Claude Code's durable, token-efficient memory: cross-platform, multi-vault setup and health checks for the Local REST API bridge (one MCP server per vault), a vault contract guard whose frontmatter and ASCII rules ship off and whose canvas shape check ships on, gardening and reflection agents, canvas and map-of-content generation, and graphify wiring into a separate codegraphs vault - no vault path hardcoded.

by Matthew Badali

Two adversarial reviewers from different model families - a Claude agent and a Codex agent - tear apart a Claude Code agent, skill or plugin against one shared rubric and return a viability verdict with cited defects, weaknesses and the edits worth making. Use when reviewing a skill, agent, command or plugin as an artifact: 'is this skill worth shipping', 'tear apart this plugin', 'will this agent's description actually trigger', 'review my SKILL.md', 'does this duplicate something we already have'. Coverage is a reported outcome: if only one reviewer runs, or the two models cannot be resolved to different families, the report says so in its own title rather than passing a single-family opinion off as two. Do NOT use for architecture, implementation plans, product ideas or code diffs - use crew's architect-reviewer, planner or code-reviewer. Standalone - its own two-model config, never crew's. No hooks.