Skip to content
Josemalyson Oliveira edited this page Jun 27, 2026 · 1 revision

FAQ

Frequently asked questions about Custom Skills.


General

What is Custom Skills?

Custom Skills is a collection of 5 production-ready skills that teach AI agents how to perform complex, multi-step tasks with high quality. Each skill is a self-contained "behavior module" with structured prompts, reference materials, and ready-to-use templates.

What AI agents are compatible?

These skills work with any LLM-based coding assistant that supports structured instructions:

  • Claude Code (Anthropic) — Native support
  • GitHub Copilot — With AGENTS.md integration
  • OpenAI Codex — With instruction files
  • Custom harnesses — Adapt the SKILL.md format

Is it free?

Yes! Custom Skills is open source under the MIT License.


Installation

How do I install a skill?

# Clone the repository
git clone https://github.com/josemalyson/custom-skills.git

# Copy a skill to your agent's directory
cp -r custom-skills/learning-trail ~/.claude/skills/

Can I install only one skill?

Yes! Each skill is independent. Install only what you need.

Where should I install skills?

For Claude Code:

~/.claude/skills/

For other agents, check their documentation for the skills directory.


Usage

How do skills activate?

Skills activate automatically based on trigger phrases in the skill's description field. When your prompt matches these phrases, the agent loads the skill automatically.

Can I modify a skill?

Yes! Each skill is a self-contained Markdown file. Edit the SKILL.md file to customize the workflow, add phases, or adjust trigger phrases.

Do skills work offline?

Most skills work offline. The HTML-first templates are fully self-contained. Some skills like k8s-investigation require external tools.

Can I use multiple skills at once?

Yes! Skills can reference and enhance each other. For example, k8s-investigation output uses html-first format.


Skills

What is agent-code-detective?

A systematic methodology for reverse-engineering any AI agent open-source project. It turns your agent into a "code detective" that can map architectures.

What is html-first?

Generates self-contained HTML artifacts for spatial, visual, and interactive tasks. Based on the thesis "HTML is the new Markdown."

What is k8s-investigation?

360-degree deep investigation for Kubernetes clusters and VPS servers. Orchestrates 12 elite tools for security, performance, and compliance.

What is learning-trail?

Generates validated learning paths with atomic resource verification. Every recommended resource is verified before inclusion.

What is pre-commit-quality-gate?

Auto-generates 14-step pre-commit pipelines for any tech stack. Shift-left quality gate that fingerprints your project.


Contributing

How do I add a new skill?

See Creating-Custom-Skills for detailed instructions.

Can I request a feature?

Yes! Use the "Feature Request" issue template on GitHub.

Can I report a bug?

Yes! Use the "Bug Report" issue template on GitHub.

Can I request a new skill?

Yes! Use the "Skill Request" issue template on GitHub.


Technical

What languages are supported?

  • Go — gofmt, golangci-lint, govulncheck, deadcode, modernize, k6
  • Python — ruff, mypy, bandit, safety, vulture, pytest, mutmut, locust
  • Node.js — prettier, eslint, typescript, knip, jest/vitest, stryker, k6
  • Java — google-java-format, checkstyle, spotbugs, OWASP, JaCoCo, pitest, gatling

What tools does k8s-investigation use?

12 elite tools:

  • Observability: Glances, Netdata, Popeye, K9s
  • Network: Kubeshark, bpftrace
  • Security: Trivy, Kubescape, kube-bench, Lynis
  • Defense: Falco, kube-hunter

How does learning-trail validate resources?

Each resource undergoes:

  1. Mandatory fetch — Read the resource
  2. Substantive extraction — Extract key concepts
  3. Validation criteria — Coverage, depth, recency
  4. Boost signals — Author reputation, venue prestige

Still Have Questions?


Last updated: June 2026

Clone this wiki locally