Skip to main content
affpapa.org
RU EN
skill Claude Code CX Codex Cursor GitHub Copilot

thermos — Claude Code skill: parallel double review

Runs two code-review subagents in parallel as background tasks, then folds their findings into one prioritised report.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add cursor/plugins --skill thermos -a claude-code
Open on GitHub

What thermos does

Thermos takes the diff for a branch or PR, hands it to two reviewers running in the background, and merges their reports into a single list. thermo-nuclear-review-subagent covers bugs, breakages, security holes, developer-experience regressions and feature-flag leaks. thermo-nuclear-code-quality-review-subagent looks at structure: file growth, spare abstractions, spaghetti, overall codebase health.

Both start in one message with run_in_background: true and see the same scoped context, so neither has to guess at the change. The synthesis step puts findings first, drops duplicates, weights whatever both reviewers flagged, and leaves disagreements to your judgement. If the separate background summaries are already on screen, the skill asks for a combined verdict rather than a full rehash.

One setting shapes how you use it: disable-model-invocation: true means the agent will not pick the skill up from its description alone. You call it by name. And when the scope stays vague, with no diff, branch or file list, the reviewers end up guessing.

What it helps with

  • audit a branch for bugs and structural rot in one pass
  • hand the same diff to two reviewers
  • merge two reports into one deduplicated list
  • catch feature-flag leaks before a merge
  • spot file growth and extra abstractions in a PR

Example prompts for the agent

  • thermos: review this branch before I merge, findings first
  • run a double review on the PR diff
  • thermos on my current changes — bugs plus structure

How to install thermos

1Skills CLI (Vercel)

npx skills add cursor/plugins --skill thermos -a claude-code

2Manually: put the files in place

~/.claude/skills/thermos/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/thermos/

From the same set18

The whole set →
unslop cursor This skill removes common AI writing patterns, filler, and vague phrasing while keeping the meaning and intended tone. Use it when polishing text. Claude Code CX Codex Cursor +1 27,700installs thermo-nuclear-code-quality-review cursor Runs a strict maintainability audit of code changes, abstractions, large files, and branching complexity. Use it when a routine review needs deeper structural scrutiny. Claude Code CX Codex Cursor +1 22,200installs deslop cursor Inspects branch diffs against main to remove AI-generated code clutter and inconsistent styling. Ideal for polishing branch changes before pull request review. Claude Code CX Codex Cursor +1 9,300installs how cursor Explores codebase architecture and runtime execution paths before you edit code. Helps determine module ownership, layering, and placement for new features. Claude Code CX Codex Cursor +1 5,900installs why cursor The skill investigates why code took its current shape by checking Git history and available MCP sources, then returns cited findings about decisions and tradeoffs. Claude Code CX Codex Cursor +1 5,700installs fix-merge-conflicts cursor Resolves merge conflicts in a branch, then checks the build and tests before wrapping up. Use it when several conflicted files need focused fixes without a refactor. Claude Code CX Codex Cursor +1 2,700installs thermo-nuclear-review cursor A skill for comprehensive analysis of branch changes: checks code for bugs, broken functionality, security vulnerabilities, and feature-gate leaks. Use when you need thorough review before merge. Claude Code CX Codex Cursor +1 2,700installs review-and-ship cursor This skill automates code review before shipping: checks for bugs, tests, and intent fit, then commits and opens a PR. Use it to verify your work is ready before merge. Claude Code CX Codex Cursor +1 2,500installs

Similar

ask Yeachan-Heo Skill to route prompts through local CLIs of different models (Claude, Codex, Gemini, Cursor and others) and save results as artifacts. Claude Code CX Codex Cursor +1 710installs salesforce-apex-quality official github Checks Apex for governor-limit risks, security gaps, and missing tests. Use it when GitHub Copilot reviews or generates Salesforce classes, triggers, and test code. GitHub Copilot 884installs difit-review yoshiko-pg Launches the local difit viewer for git diffs and attaches review notes to exact code lines. Runs only upon explicit request. Claude Code CX Codex Cursor +1 1,900installs diagramming-code trailofbits Generates Mermaid diagrams from code graphs using Trailmark. Useful for visualizing call paths, module dependencies, and data flows. Claude Code CX Codex Cursor +1 4,200installs resolve-conflicts tailcallhq Handles Git merge conflicts plan-first: audits the files, rebuilds lock files, rescues edits from deleted files, and asks you when the diff is ambiguous. Claude Code CX Codex Cursor +1 683installs react-audit-grep-patterns official github Verified grep pattern library for auditing React projects before upgrading to React 18.3.1 or 19. It spots deprecated APIs and syntax breaking changes. GitHub Copilot 854installs improve-codebase-architecture Matt Pocock This skill scans a codebase for architectural friction, then builds a visual HTML report with refactoring candidates. Use it before a deeper architecture review. Claude Code 1,163,035installs tdd Matt Pocock Guides the red-green TDD loop: ensures tests are written first, keep them at public interfaces, and avoid coupling to implementation details. Claude Code 1,141,062installs

Questions about thermos

How do I install thermos in Claude Code?

Run npx skills add cursor/plugins --skill thermos -a claude-code. The skill file lands at ~/.claude/skills/thermos/SKILL.md.

What does thermos do?

It sets the review scope, launches two subagents in the background and folds their findings into one report. Duplicates get dropped and overlapping findings carry more weight.

Why doesn't thermos fire on its own?

It ships with disable-model-invocation: true, so the agent will not trigger it from a task description. You invoke it by name in your prompt.

See also

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor