chain-of-verification — Claude Code skill: verification
chain-of-verification is a Claude Code skill that checks answers through independent questions and follow-up verification. It runs for fact-checking, technical claims, and multi-step reasoning.
no license ★ 155 updated 2026-10-29
npx skills add serpro69/claude-toolbox --skill chain-of-verification -a claude-codeWhat chain-of-verification does
The skill drafts an answer, then creates 3–5 questions aimed at testing its claims. It answers those questions separately and revises the draft when the findings disagree. The sequence matters: verification and source lookup start only after all questions have been formed.
In standard mode, the checks happen within one conversation turn. Isolated mode uses Claude Code’s Task tool to start subagents. It costs roughly 8–15 times the base token count; standard mode costs about 3–5 times. Knowledge lookup uses capy_search; capy_index and capy_fetch_and_index must not be called during the workflow. Automatic invocation is off by default.
What it helps with
- check a technical claim about an API or library
- spot errors in a chain of reasoning
- verify dates, statistics, or specifications
- review code analysis where security matters
Example prompts for the agent
- Use
$kk:chain-of-verificationto check this answer: what is the complexity of Python’ssorted()? - Verify how the authentication system works with
$kk:chain-of-verification:isolated --explore. - Check the year TCP was standardized with
$kk:chain-of-verification:isolated --haiku.
How to install chain-of-verification
1Skills CLI (Vercel)
npx skills add serpro69/claude-toolbox --skill chain-of-verification -a claude-code2Manually: put the files in place
~/.claude/skills/chain-of-verification/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/chain-of-verification/
1Skills CLI (Vercel)
npx skills add serpro69/claude-toolbox --skill chain-of-verification -a codex2Manually: put the files in place
.agents/skills/chain-of-verification/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add serpro69/claude-toolbox --skill chain-of-verification -a cursor2Manually: put the files in place
.cursor/skills/chain-of-verification/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install serpro69/claude-toolbox kodex-plugin/skills/chain-of-verification2Manually: put the files in place
.github/skills/chain-of-verification/SKILL.md1Skills CLI (Vercel)
npx skills add serpro69/claude-toolbox --skill chain-of-verification -a gemini-cli2Platform CLI
gemini skills install https://github.com/serpro69/claude-toolbox --path kodex-plugin/skills/chain-of-verificationManually: put the files in place
.devin/skills/chain-of-verification/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about chain-of-verification
How do I install chain-of-verification in Claude Code?
Run npx skills add serpro69/claude-toolbox --skill chain-of-verification -a claude-code. The skill file is placed at ~/.claude/skills/chain-of-verification/SKILL.md.
What does chain-of-verification do?
It creates questions to test an initial answer, checks them independently, then revises the answer if the findings conflict. The skill has standard and isolated modes.
How do the chain-of-verification modes differ?
Standard mode checks claims within one conversation turn and costs about 3–5 times the base tokens. Isolated mode starts subagents with Task, costs roughly 8–15 times the base tokens, and supports codebase verification.