claude

terminals

Seven verbs: /explore /converge /optimize /recommend /frame /act /hold. Turn a messy idea-pile into an answer that carries its own receipt, then act on it and keep it true.

1 on GitHubby Tej Desaiv0.2.0AGPL-3.0-or-later AND CC-BY-4.0updated 1 month ago
/

Plugin terminals from marketplace terminals in Intuition-Labs-LLC/terminals-skills. Commands read from the repo's manifest — 1 month ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use and it starts local MCP servers. The exact commands are listed below; read them before installing.

What it installs

Skills (10)

  • act

    Use when a settled answer needs to become real action and the user wants the steps put into motion through the tools they already have ("make it happen", "set this up", "do it", "get this going", "book it"). Turns the locked answer into the smallest set of concrete steps and runs them through the ho

  • converge

    Use when several competing ideas, options, or hypotheses are already on the table and a single coherent answer is needed. Collapses a noisy pile of competing ideas into one coherent answer and returns a checkable receipt (done = true / R=1), or an honest partial when it can't. Reach for this at deci

  • explore

    Use at the very start of a problem, when a question is fresh and its full space of distinct angles should be opened before anything is decided. Produces up to 7 genuinely different framings, as different as possible on purpose, each its own angle rather than a variation of one idea. Reach for this w

  • frame

    Use when the user points at their own existing context (a folder, a doc, a pile of notes, a repo) and wants to start from it rather than from scratch ("here's my stuff, take it from here", "use what's in this directory", "pick up where I left off"). Reads their context, translates it onto the 7-grid

  • glue

    Use when two or more things that each already hold need to become one, and the real question is whether they hold together ("do these two plans fit", "merge these answers", "can we run both", "join these"). Treats each as its own cluster and converges across the seam, returning one combined answer (

  • hold

    Use when a decision was settled earlier and something has changed, and the user wants to know if it still holds rather than redo it from scratch ("does this still work now that X happened", "did the plan survive", "re-check our call"). Re-converges with memory and reports which parts drifted. Use ho

  • mint

    Use when the user wants the finished thing made for them, the actual artifact written out and kept, not just options or a pick ("make me a plan", "draft the doc", "write up the decision", "give me a checklist I can use", "produce X"). Frames the intent, runs the loop (explore, converge, optimize), g

  • optimize

    Use right after a coherent answer is reached (done = true / R=1) to find its cheapest, cleanest equal form (fewer moving parts, less effort, lower noise) without breaking it. Reach for this when an answer holds together but feels heavier than it needs to be, or the user says "simplify this" / "make

  • recommend

    Use when the user asks you to decide for them, or signals they're unsure and want your call ("what should I do", "you pick", "I don't know, recommend something"). Take agency: run the whole loop yourself (explore your own best lines, converge, optimize) and return a clear pick backed by a receipt. T

  • tinker

    Use when an answer is not settled yet and the user wants to keep working it with you in the loop, pass after pass, until it holds ("let's work this", "keep at it until it's right", "iterate on this with me", "this isn't there yet, keep going"). Runs the open loop: lay the parts out, bring them toget

Slash commands (10)

  • /terminals:act

    Turn a settled answer into real next steps and run them through the tools your host already has. Acts only with your go-ahead, one step at a time.

  • /terminals:converge

    Bring a messy pile of ideas together into one coherent answer, with a receipt you can check (done = true / R=1).

  • /terminals:explore

    Open a question into its full fan of distinct angles, as different as possible, on purpose, before you pick one.

  • /terminals:frame

    Start from my stuff. Pull the context I already have (this folder, my notes) into the system and get going.

  • /terminals:glue

    Join two or more things that already hold into one. Check they hold together, and hand back the combined trinket, or name the seam that won't take.

  • /terminals:hold

    Re-check a settled answer as the world moves. Reports which parts came loose so you fix the drift instead of re-deciding the whole thing.

  • /terminals:mint

    Make me the finished thing. Take what I want, run the loop, and write out a clean finished piece I can keep and use, with a receipt that says it holds.

  • /terminals:optimize

    Take an answer that already holds together and make it the best one. Same answer, cheapest and cleanest form, without breaking done = true.

  • /terminals:recommend

    You decide for me. Run the whole loop (explore, converge, optimize) on your own best lines and hand back a pick, with the receipt.

  • /terminals:tinker

    Work it until it holds. Open the loop with you in it: lay the parts out, bring them together, find what's loose, trim or re-open just that part, and go round again until the trinket holds or you stop.

Subagents (2)

  • explorer

    Explores ONE angle of a problem deeply and reports a tight, evidence-backed case for that single angle. Spawned up to 7 times in parallel by /recommend to cover the 7-grid, where each instance owns one point.

  • referee

    Judges how well a set of explored angles fit together. Given the angles (and their cases), returns an honest pairwise coherence matrix (0..1) and reads the gate (which trios would lock, whether the set can reach done = true). Feeds the converge engine its coherence judgment.

MCP servers (1)

  • terminals

    runs: python3 ${CLAUDE_PLUGIN_ROOT}/mcp/server.py

Hooks (8)

  • SessionStart

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh SessionStart

  • UserPromptSubmit

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh UserPromptSubmit

  • PreToolUse

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh PreToolUse

  • PostToolUse

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh PostToolUse

  • Stop

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh Stop

  • SubagentStop

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh SubagentStop

  • PreCompact

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh PreCompact

  • SessionEnd

    ${CLAUDE_PLUGIN_ROOT}/hooks/terminals_hook.sh SessionEnd

View source on GitHub

Similar plugins

avoid-ai-writing

claudecodex

Audit & rewrite content to remove AI writing patterns ("AI-isms"). Supports detect-only and edit-in-place modes, voice profiles, and iterate-to-convergence.

1 skill

conorbronsdon/avoid-ai-writing 5,612

sprint

claude

Autonomous multi-agent development framework with spec-driven sprints. Write specs, run /sprint, and let coordinated agents (backend, frontend, QA, UI testing) handle implementation through self-iterating convergent loops.

4 skills · 6 commands · 9 agents

jeremylongshore/tons-of-skills-marketplace 2,917

pmo

claudecodex

Enterprise program management office: portfolio governance, program and project delivery, delivery risk, benefits realization, and adoption.

9 skills

cbrock84/headcount 2,573

pm-method

claude

Process discipline for professional work — the difference between an agent that produces output and one that produces outcomes. Interview the requester before building, brainstorm divergent-then-convergent, write and execute real plans, verify before declaring done, orchestrate subagents, work incrementally, and simplify what you built.

10 skills

mohitagw15856/pm-claude-skills 1,544

Grounds a coding agent's architecture decisions in real arXiv prior art: parallel isolated reads of papers found category-wise, converged into one recommended path with citations.

1 skill

UditAkhourii/neuroarxiv 438

dev

claude

Universal development assistant with real skill discovery. v1.31.3: Statusline extracted to standalone plugin. 46 skills.

47 skills · 15 commands · 16 agents · 1 MCP server

MadAppGang/claude-code 288

Maintain terminals?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex pluginslisted