/skill

Make it your own.

In Claude Code, use the slash command and add your context. In Codex, select skill from the just-vibe skill picker, then send the same brief.

Version 0.11.0 also supports /jv skill, /just-vibe skill and /jv:skill in Claude. See shortcut setup and context examples.

What the agent does

1. Inspect existing skills and the target host's packaging conventions; define the boundary, inputs, and one matching and one near-miss request.
2. Write actionable instructions, separate repeatable execution into utilities, and move conditional detail into references only when needed. When the project generates its skills, edit the source catalog or generator, not the generated SKILL.md.
3. Validate triggering with the matching and near-miss requests, and behavior with fixtures.

Inputs

• workflow purpose, activation conditions, target host, and examples.

Optional context: scope, references, constraints, successCriteria, environment, mode, budget.

Scope

Reads

Create/update a skill and necessary supporting utilities; installation/publication is separate unless requested.

Writes

Apply: only the requested local changes and relevant isolated verification. Inspect/plan requests remain inspection/planning. External actions require their exact action and target in session authorization.

Mode

Apply; workflow purpose, activation conditions, target host, and examples.

Prerequisites

Resolve the user brief and inspect the relevant project or supplied evidence. External capabilities are optional unless the selected action actually needs them.

Expected output

• Skill entry point and supporting assets with capability requirements, usage examples, matching boundaries and independent fixture evidence.

How the work is checked

• A matching request triggers the intended behavior; a nearby out-of-scope request does not acquire unrelated instructions.

When to stop or clarify

• Do not create universal catch-all skills, silently enable hooks, or claim executable guarantees from prose alone.

Handling missing context

Infer

Resolve the named files, existing scripts, current task and earlier corrections from the conversation and repository.

Assume

Use the narrowest interpretation that completes a reversible local task; state a consequential assumption once.

Ask

Ask when competing targets or incompatible success conditions would change the result; continue independent inspection first.

Technical guidance

Evidence

Read the intended invocation, neighboring workflows, input/output contract and required supporting assets.

Method

Keep the entry point focused and route conditional depth to self-contained references; preserve user scope and existing invocation policy.

Pitfall

Long generic checklists dilute useful detail, and a linked guide that is never loaded cannot improve behavior.

Check

Validate frontmatter/references and exercise a representative request plus a nearby nonmatching request; structural validity is not behavioral proof.

Situational decisions

When guidance merely repeats generic model capabilities: Remove it and retain decisions, invariants and examples that change behavior.

The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.

Keep exploring