CreateSkill

Scaffolds, validates, tests, and optimizes DevOS skill units end to end.

Updated Sep 3, 2026
One-click install
npx skills add https://github.com/ruban-s/DevOS --skill createskill-ruban-s
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: CreateSkill
Source: https://github.com/ruban-s/DevOS/tree/main/skills/CreateSkill
Command: npx skills add https://github.com/ruban-s/DevOS --skill createskill-ruban-s

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Hand-written skill files drift from canonical structure, fail to trigger reliably, and leak sensitive data into public releases. This Skill owns the entire skill lifecycle — scaffolding, validation, canonicalization, effectiveness testing, and trigger tuning — so every skill ships in a proven, compliant shape. ## Core Features & Use Cases - Scaffold and Update: Create new skills with canonical TitleCase naming, flat two-level trees, frontmatter with USE WHEN triggers, routing tables, gotchas, and examples; extend existing skills with new workflows or tools. - Validate and Canonicalize: Audit skills against the SkillSystem.md doctrine — naming, frontmatter format, routing integrity, hygiene gates — and repair non-compliant structures automatically. - Test and Optimize: Race with-skill agents against no-skill baselines on realistic prompts, then tune the frontmatter description with 20 should/shouldn't-fire probes to fix triggering accuracy. - Use Case: You ask to create a recipe-management skill. The Skill reads the doctrine, raises the directory with Workflows/ and Tools/, drafts SKILL.md, runs the hygiene gate, and offers a TestSkill run to prove it actually improves outcomes. ## Quick Start Ask the agent to create a new skill for a task you have in mind, or to validate, test, or improve an existing skill by name.

Frequently Asked Questions about CreateSkill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a new AI agent skill with proper structure?▼

Invoke the CreateSkill workflow, which reads the canonical SkillSystem.md doctrine, classifies the skill into one of nine types, scaffolds the directory with SKILL.md, Workflows/, and Tools/, and drafts frontmatter with USE WHEN triggers. It finishes with a ship checklist covering naming, routing, and hygiene.

How do I test whether a skill actually improves agent output?▼

Use the TestSkill workflow: draft 2-4 realistic prompts, then run paired subagents side by side — one guided by the skill, one baseline without it. Compare outputs and transcripts, then iterate through the ImproveSkill workflow if the skill shows no lift.

Why is my skill not triggering when it should?▼

The frontmatter description controls firing, and models under-call skills by default. The OptimizeDescription workflow builds 20 should-fire and should-not-fire probes, scores the current description with batched subagent judgments, rewrites the line, and re-scores until accuracy exceeds 85%.

What naming conventions do skill files require?▼

Public skills use TitleCase directories like Blogging or CreateSkill; private skills use a leading underscore with all caps like _MYSKILL. Workflow, reference, and tool files are all TitleCase (Create.md, ManageServer.ts). Kebab-case, snake_case, and all-caps names are rejected by validation.

Can a skill contain personal data or API credentials?▼

No. Both public and private skill bodies must be publish-clean; personal config, credentials, and identity data live under DEVOS/PROFILE/CUSTOMIZATIONS and are referenced by path. The SkillHygieneGate.ts tool enforces this deterministically against a canonical deny-list before any skill ships.