Agent Skills Documentation Mirror (agentskills-docs)

An offline-ready, deterministic, and verbatim knowledge base mirror of the official Agent Skills documentation suite.

Designed for dual-audience utility:

1. AI Agent Ingestion & RAG: High-integrity, deterministic markdown with zero paraphrasing, clear document boundaries, and structured metadata for agent connectivity and retrieval pipelines.
2. Human Director Orchestration: Clean, scanable architecture enabling technical directors to navigate, review, and orchestrate agent capabilities without code-reading friction.

This is a read-only AgentSkills.io documentation mirror for human-AI collaboration. Mirrored source documents remain unchanged unless the official published AgentSkills.io documentation changes. AI collaborators should read AGENTS.md for navigation, authority, progressive-disclosure, and usage rules.

---

Provenance & Verification

Parameter	Value
Source Domain	agentskills.io
Upstream Repository	agentskills/agentskills
Upstream License	Apache License 2.0
Retrieval Date	September 6, 2026
Scope Audit	9 canonical documentation pages (100% concordance across llms.txt, /sitemap.xml, and internal crawler)
Extraction Integrity	Verbatim extraction from canonical .md endpoints. Zero summarization or synthetic rewrites.

---

Documentation Catalog

Document	Title	Scope & Description	Source URL
00-index.md	Documentation Index	Master index table, extraction methodology notes, and raw upstream llms.txt.	llms.txt
home.md	Agent Skills Overview	Overview of Agent Skills framework, architecture, ecosystem, and integration principles.	/home
specification.md	Specification	Complete formal specification: directory structure, SKILL.md frontmatter, schema validation, and naming constraints.	/specification
clients.md	Client Showcase	Comprehensive registry of all 46 tools, IDEs, and agent frameworks supporting Agent Skills.	/clients
skill-creation-quickstart.md	Quickstart	Hands-on guide to creating a first skill, configuring directories, and executing within VS Code.	/skill-creation/quickstart
skill-creation-best-practices.md	Best Practices	Guidelines for scope calibration, progressive disclosure, context budgeting, and tool orchestration.	/skill-creation/best-practices
skill-creation-optimizing-descriptions.md	Optimizing Descriptions	Trigger tuning, negative constraints, synthetic evaluation queries, and prompt calibration.	/skill-creation/optimizing-descriptions
skill-creation-evaluating-skills.md	Evaluating Skills	Eval-driven development lifecycle, test case generation, execution scoring, and quality gates.	/skill-creation/evaluating-skills
skill-creation-using-scripts.md	Using Scripts in Skills	Bundling executable scripts, subprocess isolation, runtime environments, and deterministic outputs.	/skill-creation/using-scripts
client-implementation-adding-skills-support.md	Adding Skills Support	Architecture guide for agent developers implementing skill discovery, parsing, and execution runtimes.	/client-implementation/adding-skills-support

---

Agent & RAG Ingestion Guide

Document Structure Conventions

• Source Provenance Header: Every document begins with # Source: <canonical-url> on line 1, followed immediately by the verbatim # <Page Title> on line 3.
• Deterministic Content: All tables, code fences, shell scripts, and parameter lists are preserved in exact upstream form.
• Non-Content Removal Disclosure: The only text omitted from source files is the transient 3-line Mintlify index banner (> ## Documentation Index...). No substantive text, parameter definitions, or code examples were modified.
• Client Ordering: Canonical .md endpoints were utilized to eliminate non-deterministic client-side JavaScript shuffling present on browser rendering of the home and clients pages.

Programmatic Manifest (sources.json)

For automated RAG pipelines, pipeline indexing, and cache invalidation, refer to sources.json, which provides:

• Document titles and canonical source URLs
• Character counts, byte sizes, and line counts
• Local SHA-256 checksums for each normalized mirror file

---

License & Attribution

This documentation mirror is derived from the official Agent Skills specification provided by the Agent Skills community. The source materials are licensed under the Apache License 2.0. See the LICENSE file for complete license terms.