claude

conductor

Mixed-model orchestration at `high` effort: Claude for design, judgment and high-stakes writing, Codex CLI for mechanical execution, volume writing and cross-family verification.

1 on GitHubby Ivan Obothv1.18.8MITupdated 3 days ago
/

Plugin conductor from marketplace agent-ops in IvanOboth/conductor-plugin. Commands read from the repo's manifest — 2 days ago.

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

What it installs

Skills (14)

  • agent-browser

    Browser automation CLI for AI agents. Use when the user needs to interact with websites, navigate pages, fill forms, click buttons, take screenshots, extract data, test web apps, or automate any browser task. Also trigger when the user says "test this login page", "verify the signup form", "check if

  • codex-computer-use

    Ask Codex (GPT-6.1 Sol) to run local app verification that needs computer use — launching apps, driving a browser or simulator, capturing screenshots, or any independent runtime inspection outside Claude's current context. This is how GPT-6.1 Sol is invoked for computer-use verification. Use when th

  • codex-review

    Quick Codex review of recent work. Summarizes what was attempted and done, sends to Codex for independent review, and optionally fixes identified issues. Use after completing any task to get a second opinion.

  • conductor-claude

    Orchestrate a full Conductor run inside the Claude family only — Opus 5.5 and Fable 5.2 lanes, effort as the dial, context-independent adversarial review in place of the cross-family gate. Use when the Codex/ChatGPT quota is spent, Codex is unreachable, or the work must not leave the Claude family.

  • conductor-core

    Orchestrate Codex-first work with GPT-6.1 Sol owning the run, GPT-6.1 Sol as the primary worker and fresh-context reviewer, Luna as the cheap bounded tier, work-sized parallel lanes, and optional Opus design or review. Use for Conductor Core, Codex-first orchestration, or Conductor without a Fable d

  • conductor

    Coordinate mixed-model work across Claude Code and Codex. Keep the current main agent responsible for grounding, work orders, integration and final acceptance; delegate bounded execution and independent review when worthwhile. Produce readable explanatory HTML reports with evidence chosen for the ta

  • feature-map-update

    Keep a project's feature map true in the same pull request that changes the app. Runs the map checker against the branch's base, lists the feature files the diff touches and any new route no feature covers, updates those files from the changed code, and re-drives the changed recipe steps live. Use b

  • how

    Explain how a subsystem, feature or runtime flow works in this codebase, at the level of a senior engineer onboarding onto it — overview, key concepts, the flow step by step with file references, where things live, gotchas — and optionally critique the architecture with independent reviewers from di

  • journey-review

    Walk a persona's real job through a product (a salesperson visiting an outlet, taking an order, adjusting it, collecting payment, giving a receipt; a back-office user settling a truck) using the project's feature map and helper, find where the flow does not match how that person works or is not intu

  • mobile-app-testing

    Test native mobile apps on Android or iOS devices, diagnose runtime and sign-in failures, fix scoped gaps, and deliver verified evidence. Use for Expo development-client and OTA verification, APK/IPA QA, mobile release checks, or native regression testing.

  • run-report

    The closing convention for any substantial agent run (conductor, feature implementation, nightly shift, cloud routine). Use at the END of a run to publish the run report to GitHub, flip status labels, and append the telemetry ledger. Trigger when a run wraps, when the user says "close out the run",

  • verify-skill-create

    Generate a project-local verification skill (`.claude/skills/verify-<app>/`) that lets any agent drive the real app the way a user does — a small helper CLI (target or launch, doctor, sign-in as a test persona, shot, record, stop) plus a feature map that says where every user-facing feature lives, w

  • verify-skill-eval

    Measure whether a change to agent guidance (a verification skill, a feature map, an AGENTS.md section, a rewritten skill) makes agents better at real tasks — paired headless Claude trials in two checkouts, same organic prompts, neutral labels, metrics from the transcripts (time, turns, tool calls, c

  • verify-skill-maintain

    Scheduled upkeep pass that keeps a project's verification skill and feature map true to the app — checker first, one source reader per feature file, one live pass that drives every feature, then at most one PR of proven corrections. Product regressions found on the way are reported, never written in

Subagents (5)

  • bulk-lane

    Conductor's cheap Claude execution lane. Use for mechanical work with exact anchors — renames, applying a defined contract across files, small migrations — one lane per item. Since 2026-09-22 Opus 5.5 at low is the default for mechanical sweeps (Anthropic reports low comes close to Opus 5 at high on

  • design-lane

    Conductor's taste lane. Use for user-facing surfaces where "looks right" is the acceptance test — components, layout, visual grammar, copy, API shape. Pinned to Opus 5.5 at high effort (its ceiling in Conductor routing) so taste-critical work does not silently inherit a low session effort. For a cri

  • exec-lane

    Conductor's default execution lane since 2026-09-22. Use for well-specified implementation, refactors, migrations, terminal/CI/infra work, and test authoring. Pinned to Opus 5.5 at high effort (Ivan, 19 Oct 2026: implementation runs at high, not medium). Use design-lane for taste-critical surfaces,

  • verify-lane

    Conductor's judgment verification lane. Use to adversarially check a finding, a diff, or a multi-step runtime flow where "did this actually work correctly" needs judgment rather than a green console. Since 27 Sep 2026 it runs on critical changes only: anything that deploys to production or changes p

  • write-lane

    Conductor's high-stakes writing lane. Use for prose a human will read and judge the author by — a counterparty email, a proposal, an investor or board document, a script or narration, the final edit of a draft another lane produced. Pinned to `fable` at high effort (Fable 5.2 since 10 Nov 2026) so v

Hooks (2)

  • PostToolUse

    python3 "${CLAUDE_PLUGIN_ROOT}/scripts/context-watch.py" || true

  • Stop

    python3 "${CLAUDE_PLUGIN_ROOT}/scripts/context-watch.py" || true

View source on GitHub

Similar plugins

Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).

39 skills

mattpocock/skills 323,108

career-ops

claudecodex

Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications

5 skills · hooks

career-ops-hq/career-ops 82,253

Use Alibaba Open Code Review from local Codex.

2 skills

alibaba/open-code-review 50,068

Backend API design, GraphQL architecture, workflow orchestration with Temporal, and test-driven backend development

9 skills · 1 command · 8 agents

wshobson/agents 42,461

codex

claude

Use Codex from Claude Code to review code or delegate tasks.

3 skills · 8 commands · 1 agent · hooks

openai/codex-plugin-cc 35,098

superset

claude

Use Superset to spawn agents, create workspaces, manage tasks, and schedule automations from Claude Code.

14 skills · 1 command

superset-sh/superset 15,548

Maintain conductor?

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