write-skill

Guides creation and testing of Specture skills using a TDD-for-documentation discipline.

1|Updated May 1, 2026
One-click install
npx skills add https://github.com/FerEscobarDev/Specture --skill write-skill-ferescobardev
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: write-skill
Source: https://github.com/FerEscobarDev/Specture/tree/main/skills/write-skill
Command: npx skills add https://github.com/FerEscobarDev/Specture --skill write-skill-ferescobardev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI behavior skills often ship untested, so agents rationalize their way around the rules. This Skill enforces a test-driven workflow for writing or editing Specture skills, requiring a failing baseline scenario before any skill content is written. ## Core Features & Use Cases - TDD cycle for documentation: Run pressure scenarios without the skill (RED), write the minimal counter (GREEN), then close loopholes by capturing agent rationalizations (REFACTOR). - Skill authoring standards: Defines the SKILL.md frontmatter format, CSO rules for the description field, body skeleton, and language conventions (English body, Spanish user-facing messages). - Use Case: When you want to add a new Specture skill or edit an existing one, this Skill forces you to first observe an agent failing without it, then write only what addresses that failure. ## Quick Start Ask the AI to create a new Specture skill for a recurring behavior problem and have it follow the write-skill TDD process.

Frequently Asked Questions about write-skill

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

FAQPage Schema
How do I create a new Specture skill?▼

Create a directory under skills/ with a SKILL.md file containing YAML frontmatter with name and description fields. The write-skill process requires you to first run a pressure scenario without the skill, document the agent's failure, then write the minimal skill that counters it.

How to write a good skill description for Claude discovery?▼

Start the description with "Use when" followed by concrete triggering conditions like symptoms, user phrases, or file states. Never summarize the workflow in the description, because Claude will follow the summary and skip the skill body.

When should I not create a new skill?▼

Do not create skills for one-off solutions, standard practices already documented elsewhere, or project-specific conventions, which belong in .specture/conventions.md. Mechanical constraints that a regex or validation can enforce should be automated instead.

How do I test a skill before deploying it?▼

Pick a scenario where the skill should fire, run it in a fresh session without the skill loaded, and observe the agent's rationalizations. Then repeat with the skill loaded and add explicit counters for every excuse the agent used to skip the discipline.

What language should Specture skill files be written in?▼

Skill bodies are written in English for multi-model compatibility and better search optimization. User-facing messages and templates produced by the skill are written in Spanish, matching the framework user's preference.