Lesson 3.2Lesson 14 of 30

Slash commands and skills

Create and configure custom slash commands and skills

Key points

1Project commands live in .claude/commands/ and are committed, so the whole team (and CI) gets them. User commands in ~/.claude/commands/ are personal and exist only on your machine.
2A skill is .claude/skills/<name>/SKILL.md. Its frontmatter configures it: context: fork, allowed-tools, argument-hint.
3context: fork runs the skill in an isolated sub-agent context: verbose or exploratory output stays out of the main conversation and only the result comes back. The fork does not see the earlier conversation, so pass what it needs as arguments.
4allowed-tools restricts what the skill may use (e.g. Read, Grep, Glob for a read-only audit). argument-hint prompts for required parameters such as "<resource> <method>" instead of letting Claude guess.
5Skills are on-demand, task-specific workflows; CLAUDE.md holds universal standards loaded in every session. Personal variants of team skills go to ~/.claude/skills/ under a different name, so they don't override the team's.

What the exam guide checks here

Objective wording from the exam guide (English).

3.2.K1Project commands in .claude/commands/ (shared) vs user commands in ~/.claude/commands/ (personal)
3.2.K2Skills in .claude/skills/ with SKILL.md frontmatter: context: fork, allowed-tools, argument-hint
3.2.K3context: fork runs a skill in an isolated sub-agent context
3.2.K4Personal skill variants in ~/.claude/skills/ under different names
3.2.S1Create project slash commands in .claude/commands/ for the whole team
3.2.S2Use context: fork for verbose or exploratory skills
3.2.S3Restrict tools during a skill with allowed-tools
3.2.S4Use argument-hint to prompt for required parameters
3.2.S5Choose skills (on-demand workflows) vs CLAUDE.md (always-loaded standards)

Answer scenario questions on this topic. Every option comes with an explanation.