/research

Make it your own.

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

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

What the agent does

1. Turn the question into answerable subquestions or compatibility claims, and inspect the project's constraints and pinned versions.
2. Check authoritative primary documentation for each claim, recording its date and exact feature boundaries; compare evidence dates and distinguish facts from your own conclusions.

Inputs

• technical question, constraints, and the time available, which bounds how deep the research goes. Requires local evidence and current primary sources when claims depend on versions or external facts.

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

Scope

Reads

Resolve the question, not implement a preferred solution.

Writes

No source changes in inspect/plan. Save only requested planning artifacts. A separately requested repair uses the relevant implementation workflow.

Mode

Inspect; technical question, constraints, and the time available, which bounds how deep the research goes. Requires local evidence and current primary sources when claims depend on versions or external facts.

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

• Claim/source/version table with citations, the alternatives considered, and a recommendation conditional on unresolved facts and compatibility assumptions.

How the work is checked

• Conflicting documentation is surfaced; an unsupported feature is not recommended based on outdated examples.

When to stop or clarify

• If authoritative information is unavailable, bound confidence and propose a small validation experiment instead of fabricating certainty.

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

Resolve the precise question, installed versions, decision constraints and primary source availability.

Method

Search authoritative version-relevant material, distinguish quoted facts from inference and compare conflicting evidence.

Pitfall

Current documentation may describe a different major than the project; search snippets alone can omit decisive conditions.

Check

Link claims to supporting passages and identify the uncertainty that remains relevant to the decision.

Situational decisions

When authoritative sources conflict or describe another version: State the conflict and testable assumption; do not combine incompatible APIs into one recommendation.

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