Skip to content
Development
Command

/isolated

Apply Chain-of-Verification (CoVe) with true factored verification using isolated sub-agents.

From plugin
claude-toolbox
1474 skills8 agents4 commands1 MCP
Install
$ npx -y skills add serpro69/claude-toolbox --agent claude-code

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/isolated

Context preview

What this command does when you run it.

Apply Chain-of-Verification (CoVe) with true factored verification using isolated sub-agents.

Command definition

isolated.md

Apply Chain-of-Verification (CoVe) with true factored verification using isolated sub-agents.

This mode spawns separate sub-agents for each verification question, ensuring they have zero context about the initial answer. Use when maximum accuracy is required.

Arguments: $ARGUMENTS

Flags

| Flag | Effect | Use Case | |------|--------|----------| | (none) | `general-purpose` agent | Default, full tool access | | `--explore` | `Explore` agent | Codebase-related verification | | `--haiku` | Use haiku model | Faster/cheaper verification | | `--agent=<name>` | Custom agent type | User-defined agents |

**Flag Parsing Rules:**

  • Flags must appear before the question
  • Flags start with `--`
  • `--agent=value` format for custom agents
  • Unknown flags are treated as part of the question
  • Flags can be combined in any order

Invocation

**If $ARGUMENTS is provided:** 1. Extract any flags from the beginning of $ARGUMENTS 2. Apply CoVe isolated verification to the remaining question 3. Pass extracted flags to the workflow

**If $ARGUMENTS is empty:** Apply CoVe isolated verification to the previous response in the conversation.

Process

Invoke the `kk:chain-of-verification` skill using the `chain-of-verification-isolated.md` workflow:

1. Generate Initial Answer (or use previous response if verifying) 2. Create 3-5 Verification Questions targeting potential errors 3. Spawn isolated sub-agents to answer each question (parallel execution) 4. Reconcile findings and produce final verified answer with agent metadata

Examples

Basic isolated verification:

/kk:chain-of-verification:isolated What is the default port for PostgreSQL?

Codebase verification with Explore agent:

/kk:chain-of-verification:isolated --explore How does the auth middleware work in this project?

Faster verification with haiku model:

/kk:chain-of-verification:isolated --haiku What is the speed of light in m/s?

Custom agent type:

/kk:chain-of-verification:isolated --agent=general-purpose What is the capital of France?

Combined flags:

/kk:chain-of-verification:isolated --haiku --explore What testing pattern does this codebase use?

Verify previous response:

/kk:chain-of-verification:isolated

(Use after receiving a response to verify it with isolated sub-agents)

Read more
Ships withclaude-toolbox

<div align="center" claude-toolbox is a collection of "tools" for all your agentic workflows — pre-configured MCP servers, skills, sub-agents, commands, hooks, statuslines with themes, and more - everything you need for AI-powered development workflows, used

Get the whole plugin, auto-invoked
Stats
147
Stars
0
Views
31
Forks
Active
Maintenance
Shell
Language
1d ago
Last commit
9mo ago
Created

Repo: serpro69/claude-toolbox