agentleFS
WillInvestWillInvest/ClaudeX/skills/think/SKILL.md

think

Skill4 starsChanged 6 months agoMITView on GitHub

Brainstorm a feature or design with an automatic Codex second opinion on every clarifying question. Use when the user runs /claudex:think, or asks for Claude+Codex collaborative brainstorming. Wraps superpowers:brainstorming.

30 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use hunt, read and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • ninehillsSkill

    hunt

    Finds root cause before applying fixes for errors, crashes, regressions, failing tests, broken behavior, and screenshot-reported defects. Use when users report in any language errors, crashes, broken behavior, regressions, failing tests, screenshot evidence, or something that used to work and now fails. Not for code review or new features.

    Also in 5 projects that use this

  • MicrosoftSkill

    read

    Read pages from a document. Returns MinerU markdown (preferred) or PyPDF text, with optional page screenshots and sub-image fetching. When markdown is available, a figure_images_meta catalog is included — use the figures parameter to fetch specific sub-image pixels. Triggers include "read page 5", "what does page 12 say", "show me the chart on page 3", "extract pages 3-5".

    Also in 6 projects that use this

  • DockerSkill

    write

    Write or edit reader-facing technical prose for immediate comprehension. Use for documentation, PR titles and descriptions, release notes, design documents, user-facing explanations, and substantive comments. Also use when asked to make writing clearer, more natural, less AI-generated, or easier to read. For documentation fixes, handles edits, formatting, self-review, and commits after research identifies what to change. Do not use for code-only tasks with no prose deliverable.

    Also in 7 projects that use this

  • genomewalkerSkill

    health

    Soul system health check with remediation

    Also in 5 projects that use this

  • kt057Skill

    check

    lint, typecheck, test を一括実行して結果をサマリーする

    Also in 6 projects that use this

  • OpenAISkill

    transcribe

    Transcribe audio files to text with optional diarization and known-speaker hints. Use when a user asks to transcribe speech from audio/video, extract text from recordings, or label speakers in interviews or meetings.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 0 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 25 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When the user runs /claudex:think, a UserPromptExpansion hook arms the gate for this session automatically (it runs in hook context, so it …

    line 12
  2. Very specific:

    Add My take: — 1-3 sentences: do you agree?

    line 20
  3. Very specific:

    Wrap up: When you transition to writing-plans, the gate auto-disarms (a PreToolUse Skill hook handles it).

    line 24
  4. Very specific:

    Fail-open by default: if Codex is unavailable or times out, the question proceeds with a note that the second opinion was skipped — mention …

    line 49

“Common” is measured: kept by at least 6 in 10 of the 25 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

claudex:think — brainstorming with a Codex second opinion

Brainstorm normally, but every clarifying question is routed through Codex for a second opinion before the user sees it. Enforcement is automatic — you never call Codex yourself, and you do NOT need to run any shell command to turn it on.

How arming works (automatic)

When the user runs /claudex:think, a UserPromptExpansion hook arms the gate for this session automatically (it runs in hook context, so it isn't subject to the permission classifier). You don't touch it.

Steps

  1. Brainstorm. Invoke Skill(superpowers:brainstorming) and follow it normally. Ask clarifying questions via AskUserQuestion as usual — the hook intercepts each one.
  2. Handle the Codex critique. When you try to ask a clarifying question, the hook blocks it. The (red) denial reason is kept short on purpose — it points you to a file (~/.claude/claudex/state/<session>-critique.md) holding your drafted question + Codex's critique, so the critique isn't shown red and then repeated. Read that file, then in ONE message:

    • Show the question you originally drafted verbatim under an **Original question:** heading (keep your recommended option), so the user can see your framing before Codex weighed in.
    • Present Codex's critique verbatim under a **Codex says:** heading.
    • Add **My take:** — 1-3 sentences: do you agree? Fold any fixes you accept into the question.
    • Re-issue the AskUserQuestion (revised if you agreed) — the revised question is the one the user actually answers. It passes through this time (one re-issue is allowed within ~2 minutes of the block).

    So each round the user sees, in order: Original question (with your recommendation) → Codex says → My take → the answer window — with no red duplicate of the critique. (If the file can't be written, the hook falls back to putting the critique inline in the denial reason.)

  3. Wrap up. When you transition to writing-plans, the gate auto-disarms (a PreToolUse Skill hook handles it); it also auto-expires after the TTL.

Design before implementation (enforced by the hook, not just by this text)

While the gate is armed you are in design mode. PreToolUse hooks block — not just discourage — these until the design is approved:

  • Editing code. Edit/Write/MultiEdit/NotebookEdit to any non-markdown file is denied (.mdx counts as code). Don't fight it: it means you skipped to implementation. Go back and brainstorm.
  • Editing instruction files. CLAUDE.md, SKILL.md, AGENTS.md, GEMINI.md are denied even though they're markdown — changing them mid-session could weaken the gate itself.
  • Implementation skills. test-driven-development, executing-plans, subagent-driven-development, frontend-design, mcp-builder are denied.
  • Shell writes. Bash commands that write code (echo >, sed -i, tee, node -e, git apply/checkout, cp/mv, …) or that touch the gate marker are denied. Read-only exploration passes. Don't try to route around the edit gate through the shell.

What you can do while armed: write the design/spec/plan markdown doc, ask AskUserQuestion, and (after you've brainstormed and the user approves) invoke writing-plans — which disarms the gate and reopens everything for implementation. So the only sanctioned path out of design mode is: brainstorm → approved design → writing-plans. If the user explicitly tells you to skip the design process, that's their call — point them to the kill switch below; don't disarm the gate yourself (the kill switch is for the user to type with a leading !).

If you never see a Codex says: block on your questions, the gate didn't arm (e.g. the plugin wasn't reloaded after install). Tell the user to run /reload-plugins and re-run /claudex:think, rather than trying to write the marker yourself (a model Bash write to ~/.claude is blocked in auto mode).

Kill switch

  • Stop early (user runs this): ! rm -f ~/.claude/claudex/armed
  • Globally off: set {"enabled": false} in ~/.claude/claudex/config.json.
  • The gate also auto-expires after ttlHours (default 6).

Behavior notes

  • Implementation gate: on by default; set {"enforceImplementationGate": false} in ~/.claude/claudex/config.json to turn off the code/skill blocking described above (the Codex question gate is unaffected). It is fail-open — a hook error never blocks an edit.
  • Read-only: Codex cannot edit files while critiquing.
  • Fail-open by default: if Codex is unavailable or times out, the question proceeds with a note that the second opinion was skipped — mention it to the user. Set {"failClosed": true} in ~/.claude/claudex/config.json to block instead.
  • This plugin's hook runs on node. Errors like "Bun not found" come from a different plugin, not claudex.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.