How do I install the Skill Authoring skill?

Run npx skillmds@latest add theopenco/skill-authoring in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Skill Authoring safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Skill Authoring?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Skill Authoring free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Skill Authoring?

theopenco (@theopenco) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Skill Authoring by theopenco — Agent Skill

Name: Skill Authoring
Author: theopenco

Skill Authoring

Create, update, and audit repository-local agent skills against the current LLM Gateway codebase. Use when adding or editing files under .agents/skills or .claude/skills, reviewing a skill for stale instructions, or checking that skill paths, commands, cross-references, and workflows actually exist.

theopenco bba24f7 3 files · 7.0 KB Updated Sep 17, 2026

File contents

Skill authoring

Write only instructions that are necessary and proven. Treat prompts, old skills, and remembered workflows as leads, not evidence.

Establish the source of truth

Read AGENTS.md and the files that implement the workflow.
Inventory both .agents/skills and .claude/skills, including hidden files and symlink targets. Keep the implementation in .agents/skills; expose every skill to Claude with a symlink.
Classify every operational claim before retaining it:
- Repository fact: prove it with the current tree using rg or by reading the implementation.
- Repository command: prove the script and package exist, then run a safe help or validation command when flags or behavior matter.
- External API or product fact: use current authoritative documentation or a live non-destructive probe.
- Machine capability: check it locally and state it as a precondition; never present one machine's installed tools or authentication behavior as a universal fact.

If a claim cannot be verified, remove it or label the exact check the agent must perform at runtime. Never invent a missing skill, tool, port, command, or path to bridge a workflow gap.

Create or edit the skill

Keep YAML frontmatter to name and description. Put all trigger phrases in the description.
Keep the body concise and imperative. Link to AGENTS.md instead of copying large procedures that will drift.
Use exact repository paths only after resolving them. Use placeholders only where the agent must choose a value.
Keep cross-skill references repository-local unless the dependency is optional and checked at runtime.
Add scripts only for repeated deterministic work. Run every added script.
Remove unused references and assets.

Validate

Run the repository skill audit from the repository root:

node .agents/skills/skill-authoring/scripts/check-skills.mjs

Then run the standard skill validator for each new or structurally changed skill when it is available, followed by pnpm format and the build required by AGENTS.md.

Review the final diff once more. Each instruction should have a concrete source: repository code, a command result, authoritative documentation, or an explicit runtime precondition.

---name: skill-authoringdescription: Create, update, and audit repository-local agent skills against the current LLM Gateway codebase. Use when adding or editing files under .agents/skills or .claude/skills, reviewing a skill for stale instructions, or checking that skill paths, commands, cross-references, and workflows actually exist.---# Skill authoringWrite only instructions that are necessary and proven. Treat prompts, old skills,and remembered workflows as leads, not evidence.## Establish the source of truth1. Read `AGENTS.md` and the files that implement the workflow.2. Inventory both `.agents/skills` and `.claude/skills`, including hidden files and symlink targets. Keep the implementation in `.agents/skills`; expose every skill to Claude with a symlink.3. Classify every operational claim before retaining it: - Repository fact: prove it with the current tree using `rg` or by reading the implementation. - Repository command: prove the script and package exist, then run a safe help or validation command when flags or behavior matter. - External API or product fact: use current authoritative documentation or a live non-destructive probe. - Machine capability: check it locally and state it as a precondition; never present one machine's installed tools or authentication behavior as a universal fact.If a claim cannot be verified, remove it or label the exact check the agent mustperform at runtime. Never invent a missing skill, tool, port, command, or path tobridge a workflow gap.## Create or edit the skill- Keep YAML frontmatter to `name` and `description`. Put all trigger phrases in the description.- Keep the body concise and imperative. Link to `AGENTS.md` instead of copying large procedures that will drift.- Use exact repository paths only after resolving them. Use placeholders only where the agent must choose a value.- Keep cross-skill references repository-local unless the dependency is optional and checked at runtime.- Add scripts only for repeated deterministic work. Run every added script.- Remove unused references and assets.## ValidateRun the repository skill audit from the repository root:```bashnode .agents/skills/skill-authoring/scripts/check-skills.mjs```Then run the standard skill validator for each new or structurally changed skillwhen it is available, followed by `pnpm format` and the build required by`AGENTS.md`.Review the final diff once more. Each instruction should have a concrete source:repository code, a command result, authoritative documentation, or an explicitruntime precondition.

Frequently asked questions