The loaditout-mcp-server lets you discover, evaluate, and install AI agent skills from a registry of 20,000+ MCP servers, with built-in security grading, memory, and human approval workflows.

Search & Discovery

• search_skills — Search by keyword with optional filters for skill type and agent platform

• smart_search — Personalized search applying your history and preferences automatically

• list_categories — Browse available skill categories with descriptions and counts

• recommend_skills — Get tailored recommendations based on your current project context

• check_capability_gap — Identify which skills you're missing to complete a specific task

Skill Evaluation & Security

• get_skill — Retrieve full details, security grade (A/B/C/F based on 7 safety criteria), and install config

• validate_action — Check if an action is safe before executing it (injection detection, freshness check)

• flag_skill — Report problematic, broken, or unsafe skills to the community

Installation

• install_skill — Get platform-specific install configuration for a single skill

• install_batch — Install up to 20 skills at once

• install_pack — Install an entire curated agent pack (e.g., research-agent, devops-engineer)

Memory & Context

• save_memory — Persist key-value data (installed skills, preferences, context) across sessions

• recall_memory — Retrieve previously saved memories, optionally filtered by type

Permissions & Human Approval

• request_permission — Ask the human owner for approval before installing a skill

• check_permission — Poll the status (pending/approved/denied) of a permission request

Usage Tracking & Proof

• report_skill_usage — Report whether a skill worked correctly to improve quality scores

• list_my_proofs — View your verified skill usage history

• verify_proof — Verify a specific execution proof by ID

• share_loadout — Share your public skill configuration, trust score, and stats