agentic-mcp-server-builder

Agent Skills Hub

Production-ready skill packs for AI coding agents: Codex, Claude Code, Cursor, Aider, Continue, and Agent Skills-compatible runtimes.

This repository focuses on deterministic skill contracts, strong validation, and reusable automation patterns for:

• agentic workflows and MCP servers
• web engineering pipelines
• ML/DL experiment workflows
• cybersecurity triage and authorized penetration-testing workflows

Why This Repository

• Standardized skill contract (SKILL.md, agents/openai.yaml, script CLI contract)
• Cross-platform packaging and compatibility checks
• Deterministic dry-run behavior for safe automation
• Scenario tests validating end-to-end skill behavior
• Security-first controls in pentest skills (scope gating and authorization requirements)

Repository Layout

./
  .github/
    workflows/
  skills/
  tests/
    smoke/
    scenarios/
    security/
  tools/

Skill Packs

This repository ships 44 production-ready skills:

• 1 meta skill: skill-creator-pro
• 43 domain skills across cybersecurity, AI/ML+DL, agentic AI, daily automation, web engineering, and authorized pentest operations

Web Builder Skills

• web-stack-planner
• web-ux-architect
• web-frontend-designer
• web-backend-builder
• web-auth-integrator
• web-database-validator
• web-api-tester
• web-frontend-tester
• web-security-auditor
• web-deploy-launcher

Pentester Skills

• pentest-engagement-planner
• pentest-recon-osint
• pentest-network-scanner
• pentest-vuln-analyzer
• pentest-web-app-attacker
• pentest-api-attacker
• pentest-auth-bypass
• pentest-injection-engine
• pentest-network-exploiter
• pentest-priv-escalation
• pentest-active-directory
• pentest-cloud-auditor
• pentest-container-k8s
• pentest-wireless-attacker
• pentest-social-engineer
• pentest-mobile-auditor
• pentest-lateral-movement
• pentest-c2-operator
• pentest-data-exfil-tester
• pentest-redteam-ops
• pentest-report-generator
• pentest-remediation-validator
• nmap-pentest-scans

Skill Contract

Each skill follows:

• SKILL.md with YAML frontmatter (name, description)
• agents/openai.yaml with display_name, short_description, default_prompt
• optional scripts/, references/, assets/

All executable scripts follow this CLI contract:

• --input <path>
• --output <path>
• --format json|md|csv (default json)
• --dry-run

Pentest skill scripts also enforce:

• --scope <scope.json>
• --target <target>
• --i-have-authorization for non-dry-run mode

JSON output contract:

• status: ok|warning|error
• summary
• artifacts
• details

Quick Start

Requirements:

• Python 3.10+
• pyyaml
• skills-ref

Install dependencies:

python -m pip install --user pyyaml skills-ref

Run full checks:

tools\run_checks.bat

Run key checks manually:

python tests\smoke\validate_all_skills.py
python tests\smoke\validate_web_builder_skills.py
python tests\smoke\validate_pentest_skills.py
python tests\smoke\run_smoke.py
python tests\scenarios\test_web_builder_skills.py
python tests\scenarios\test_pentest_skills.py
python tests\scenarios\run_agentic_scenarios.py
python tests\scenarios\run_agentskills_reference_checks.py
python tests\security\run_security_regressions.py

CI

GitHub Actions validates:

• all skill metadata and contracts
• smoke tests and scenario tests
• security regression checks
• packaging compatibility for Claude uploads

Releases and Changelog

• See CHANGELOG.md for version history.
• GitHub releases are generated from tags and curated release notes.
• Release configuration lives in .github/release.yml.
• Current release notes source: releases/v3.0.1.md

SEO and Discoverability Notes

For repository discoverability in GitHub search, keep:

• concise keyword-rich repository description
• clear README headings and domain terms (AI agents, MCP, pentest skills, web builder skills)
• active releases with tags and release notes
• consistent labels and issue templates

Recommended GitHub repository topics:

• ai-agents
• codex
• claude-code
• agent-skills
• mcp
• prompt-engineering
• developer-tools
• security-automation
• penetration-testing
• owasp
• mitre-attack
• devsecops

Compatibility

• Codex: https://developers.openai.com/codex/configuration#skills
• Claude custom skills: https://support.anthropic.com/en/articles/12522847-building-custom-skills-for-claude-ai
• Agent Skills spec: https://agentskills.io/spec

Contributing

See CONTRIBUTING.md.

Security Scope

Cybersecurity and pentest skills are for authorized use only.
Unauthorized access testing is illegal. Use dry-run mode unless written authorization exists.