writing-skills

Create and test agent skills using test-driven development methodology.

Updated Aug 28, 2026
One-click install
npx skills add https://github.com/epicodic/drift --skill writing-skills-epicodic
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: writing-skills
Source: https://github.com/epicodic/drift/tree/main/.agents/skills/writing-skills
Command: npx skills add https://github.com/epicodic/drift --skill writing-skills-epicodic

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Writing documentation for AI agents often fails because authors never verify whether agents actually follow the instructions under pressure. This Skill applies test-driven development to skill authoring so every skill is validated against real agent behavior before deployment. ## Core Features & Use Cases - TDD-Based Skill Authoring: Run baseline pressure scenarios without the skill (RED), write the skill addressing observed failures (GREEN), then close rationalization loopholes (REFACTOR). - Discovery Optimization: Guidance for writing descriptions, keywords, and naming so future agents can find the skill via search. - Testing Methodology: Detailed reference for building multi-pressure scenarios with subagents, capturing rationalizations verbatim, and bulletproofing discipline-enforcing skills. - Use Case: You want to create a new skill enforcing code review before commits. You first run scenarios where an agent is tempted to skip review, document its excuses, then write the skill with explicit counters to each rationalization and re-test until compliance holds. ## Quick Start Ask the agent to help you create a new skill using the writing-skills TDD workflow, starting with a baseline pressure scenario before drafting any content.

Frequently Asked Questions about writing-skills

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

FAQPage Schema
How do I create a new agent skill with SKILL.md?▼

Create a directory containing a SKILL.md file with YAML frontmatter defining name and description, followed by markdown instructions. Follow the RED-GREEN-REFACTOR cycle: run baseline scenarios first, write the skill addressing observed failures, then close loopholes.

How to test whether an AI agent follows a skill?▼

Use pressure scenarios with subagents that combine three or more pressures like time limits, sunk cost, and authority. Run the scenario without the skill to capture baseline rationalizations, then re-run with the skill to verify compliance.

What should go in a skill description field?▼

The description should start with "Use when..." and list only triggering conditions, symptoms, and situations in third person. Never summarize the skill's workflow, because agents may follow the description instead of reading the full skill body.

When should I not create a skill?▼

Skip skill creation for one-off solutions, project-specific conventions better placed in CLAUDE.md, well-documented standard practices, and mechanical constraints that can be enforced with regex or validation scripts instead of documentation.

Why do agents ignore skills under time pressure?▼

Agents rationalize violations using excuses like sunk cost, pragmatism, or spirit-versus-letter arguments. Counter this with explicit negations of each loophole, rationalization tables, red flag lists, and foundational principles stated early in the skill.