01
Agent Skills Overview
Framework and ecosystem overview for Agent Skills.
Documentation mirror · local knowledge source
OnTheLLow’s preserved mirror of official AgentSkills.io guidance, built for people and AI collaborators.
A preserved, locally clonable mirror of the AgentSkills.io documentation. It keeps the source material easy to browse for humans and easy to navigate for agents helping with Agent Skill creation, evaluation, optimization, scripting, and integration work.
What this repository is
A repository-shaped documentation source pack.
agentskills-docs turns the published AgentSkills.io documentation into a clean repository that can be cloned, searched, read, and supplied to AI collaborators without changing the underlying guidance.
The mirrored documents are preservation-bound reference material. The repository adds navigation and provenance around them, not a competing version of the Agent Skills standard.
Not the official development repository.
The Agent Skills standard is developed upstream in agentskills/agentskills. This project mirrors the published documentation for knowledge access and collaboration.
Why it exists
Keep one trustworthy source available across human and agent workflows.
AgentSkills.io is written for human developers. This repository preserves that same guidance in a form that can also be handed directly to AI collaborators working on Agent Skills.
Browse it
Humans get a simple index, readable Markdown, source links, and clear provenance.
Clone it
The source pack can live locally as part of a larger knowledge base and remain useful without a website workflow.
Give it to agents
AI collaborators can follow the repository’s routing and authority rules instead of relying on partial recollection or ad hoc summaries.
Documentation coverage
The complete nine-page AgentSkills.io documentation set in the current snapshot.
Each entry links to the preserved repository copy on main and the corresponding official published page.
02
Specification
The normative local authority for Agent Skills format requirements.
03
Client Showcase
Published directory of tools and platforms supporting Agent Skills.
04
Quickstart
Step-by-step introduction to creating a first Agent Skill.
05
Best practices for skill creators
Guidance on scope, calibration, progressive disclosure, and task execution.
06
Optimizing skill descriptions
Guidance for making skill descriptions trigger reliably on relevant prompts.
07
Evaluating skill output quality
Eval-driven methods for testing and improving skill output quality.
08
Using scripts in skills
Guidance for commands, bundled executable scripts, and runtime environments.
09
How to add skills support to your agent
Implementation guidance for adding Agent Skills support to an agent or development tool.
How to use it
Browse when you need context. Clone when you need a durable local source.
Browse the repository
Start with the documentation index for a human-readable map of the source pack.
Clone it locally
Use the repository directly in your local knowledge-base structure.
git clone https://github.com/Justinhubbard37/agentskills-docs.gitRoute AI collaborators through the repository
AGENTS.md tells arriving agents what is authoritative, what to read for a task, and what source files must remain unchanged.
Source integrity & provenance
Published guidance stays traceable to the official source.
The repository records where each document came from, how the local mirror was normalized, and which upstream project owns the standard itself.
Published documentation
AgentSkills.io
The canonical published documentation used to build the clean mirror.
Open AgentSkills.io
Official development repository
agentskills/agentskills
The upstream repository used as the source-history and change signal for the standard.
Open upstream repository
Mirror metadata
sources.json
Machine-readable source URLs, capture metadata, normalization rules, and local file hashes.
Inspect provenance manifestHow the mirror preserves source integrity
The repository uses canonical AgentSkills.io Markdown rather than rendered HTML. The local files add a # Source: provenance header and remove the non-substantive Mintlify AI-navigation banner. The manifest records that substantive content is not rewritten.
The mirrored corpus changes only after a verified official documentation change. The repository’s synchronization workflow checks upstream and prepares changes for human review rather than silently merging into main.
How the repository is organized for humans and agents
00-index.md is the human-readable document map. AGENTS.md is the agent navigation and authority contract. sources.json is the provenance manifest. The nine mirrored Markdown files remain the preserved source corpus.
This GitHub Pages site is only the human-facing presentation layer. It is intentionally kept separate from the knowledge-source content on main.
Preservation & freshness
Preserve the mirror. Review the change.
Mirrored documentation is expected to remain unchanged until the corresponding official AgentSkills.io documentation changes. When upstream changes are detected, the repository is designed to prepare a reviewable update rather than mutate the canonical branch silently.