Agent Skills Hub

looped-builder

Scaffold complex, multi-step coding tasks into actionable implementation plans and execute them autonomously using a Claude-driven bash loop.

wrap

Automated session cleanup and documentation tool. Proactively updates CLAUDE.md, detects automation patterns, extracts insights, and organizes pending tasks.

github-navigator

GitHub operations via gh CLI. Use for repository inspection, issues, PRs, releases, and deep codebase analysis including cloning for architectural insights.

code-reviewer

Conduct automated code reviews for local changes or remote GitHub Pull Requests. It analyzes code for correctness, maintainability, and standards using git and gh CLI integration.

ReasoningBank Intelligence

Implement adaptive learning with ReasoningBank for pattern recognition, strategy optimization, and continuous improvement in AI agents.

hs

Pre-execution security guardrails for AI agents. Validates shell commands and file reads against 400+ security patterns to block destructive operations, credential theft, and unauthorized system access.

start

Initiates automated reverse engineering by discovering codebase architecture, layers, and technology stacks to facilitate system modernization or documentation.

dyad:swarm-pr-review

Collaborative PR review using a swarm of three specialized AI agents (Correctness, Health, UX) that discuss findings and reach consensus before posting a structured summary with inline comments.

tzurot-council-mcp

Multi-perspective AI consultation for technical architecture, complex refactoring, and structured debugging.

ghost-scan-secrets

AI-powered secrets and credentials scanner. Detects hardcoded API keys, passwords, and sensitive data in your codebase with contextual analysis to reduce false positives.

executing-plans

Execute implementation plans in separate sessions with review checkpoints, ensuring task-by-task verification and robust code quality.

verification-before-completion

Enforces a strict evidence-before-assertion protocol for coding agents, requiring fresh command-line verification output before any claim of completion, success, or bug fixes.