This week's quiz focuses on Agent Skills, the emerging specification for packaging reusable capabilities that coding agents can discover and execute. Five questions covering the basics: required files, directory structure, how agents load skills, naming rules, and path handling. Answers and explanations follow the last question.
Questions
Question 1: What is the minimum required file structure for a valid Agent Skill?
A) A directory containing SKILL.md and README.md
B) A directory containing only SKILL.md
C) A directory containing SKILL.md and a scripts/ folder
D) A directory containing config.yaml and SKILL.md
Question 2: Which of the following is NOT a standard optional subdirectory in the Agent Skills specification?
A) scripts/
B) references/
C) hooks/
D) assets/
Question 3: In the progressive disclosure model, when does an agent load the full SKILL.md body?
A) At startup, along with all other skills
B) When the skill is activated
C) Only when explicitly requested by the user
D) After all reference files are loaded
Question 4: Which of the following is a valid skill name?
A) pdf_processing
B) pdfTools
C) code.review
D) code-review
Question 5: How should you reference other files within your skill?
A) Use absolute paths from the system root
B) Use relative paths from the project root where the skill is installed
C) Use relative paths from the skill root
D) Use path aliases defined in the frontmatter metadata
Answers and Explanations
Question 1: Correct Answer is B.
A skill only needs a SKILL.md file to be valid.
Question 2: Correct Answer is C.
The spec defines three optional directories. scripts/ holds executable code agents can run. references/ stores additional documentation loaded on demand (like REFERENCE.md or domain-specific files). assets/ contains static resources: templates, images, schemas, lookup tables. There's no hooks/ directory in the spec, though the name sounds plausible if you're used to git or npm conventions.
Question 3: Correct Answer is B.
Progressive disclosure is about efficient context usage. Metadata (~100 tokens) loads at startup for all skills, the full instructions load when activated, and reference files load only when needed. This pattern keeps context lean until we actually need the details.
Question 4: Correct Answer is D.
Skill names can only contain lowercase letters, numbers, and hyphens. No underscores, no dots, no camelCase. The name must also not start or end with a hyphen, and no consecutive hyphens are allowed.
Question 5: Correct Answer is C.
Use relative paths from your skill root (e.g., scripts/extract.py or references/REFERENCE.md), not from the project where the skill gets installed. This keeps the skill self-contained and portable. The spec also recommends keeping references one level deep from SKILL.md and to avoid deeply nested reference chains.
