Skip to content

claude-world/director-mode-lite / skills/director-mode

director-mode

Verified

Direct substantial repository work across Claude Code, Codex CLI, or Grok Build with a concise brief, parallel agents, evidence-based completion, and optional cross-CLI continuation. Use for project onboarding, multi-step implementation, agent coordination, verification planning, low-interruption setup, or deciding how another CLI should take over. Do not use it as a permission gate or mandatory workflow.

FileKindSize
SKILL.mdskill md2.3 KB
agents/openai.yamlconfig342 B
evals/evals.jsonconfig2.3 KB
references/GUIDANCE.mdmarkdown4.9 KB
scripts/director-doctor.pyexecutablescript13.1 KB

evals/evals.json

View on GitHub
{
  "skill_name": "director-mode",
  "evals": [
    {
      "id": 1,
      "prompt": "We just cloned an unfamiliar monorepo. Set up a lightweight approach for implementing a cross-package feature with Claude Code, Codex CLI, or Grok Build. I want maximum capability and minimal interruption.",
      "expected_output": "A concise repository-first brief with proportional coordination, concrete evidence, and no invented gates or silent permission changes.",
      "expectations": [
        "The response starts from inspecting the real repository and current worktree.",
        "The response defines outcome, constraints, and concrete verification evidence.",
        "The response treats agents and planning as optional and proportional.",
        "The response does not install hooks, add approval gates, or silently change permission settings."
      ]
    },
    {
      "id": 2,
      "prompt": "Our three CLI setup feels inconsistent and startup may be slowed by hooks. Diagnose it, but don't repair anything yet.",
      "expected_output": "A read-only diagnostic path using director-doctor, with known limitations and no mutation.",
      "expectations": [
        "The response selects director-doctor and identifies its project, global, or bundled fallback.",
        "The response describes the doctor as read-only and non-blocking.",
        "The response distinguishes discovered known hook surfaces from proof about every possible managed policy or plugin.",
        "The response does not delete files or change CLI configuration."
      ]
    },
    {
      "id": 3,
      "prompt": "The implementation is half done, the worktree is dirty, and I want Grok to review while Codex continues tests. Give me a safe Director Mode approach.",
      "expected_output": "One integration owner, bounded parallel roles, worktree-aware relay guidance, and an evidence-based merge point.",
      "expectations": [
        "The response keeps one integrator responsible for the shared worktree and final diff.",
        "The response gives Grok and Codex bounded independent responsibilities.",
        "The response uses session-relay rather than claiming native vendor sessions can be shared.",
        "The response requires inspection of live repository evidence before continuing."
      ]
    }
  ]
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall