claude

team

Envoy-and-orchestrator workflow for Claude Code with Herdr: one envoy session you talk to, an orchestrator that runs role agents in panes, briefs as files, open questions as cards, a numbered decisions file, reports by hook. Kernel only; each project adds a small overlay.

0 on GitHubby Alfred Brosev0.11.4updated 5 days ago

Not in a marketplace yet, so it loads for one session with --plugin-dir.

Plugin team in abrose/claude-plugins. Commands read from the repo's manifest — 5 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 (5)

  • team-orchestration

    The orchestrator-plus-team-agents protocol. The human talks to an envoy session, the orchestrator runs role agents in Herdr panes, briefs as files, open questions as cards, a numbered decisions file, reports by hook. Preloaded into every team agent, the orchestrator and the envoy.

  • team-role-envoy

    Envoy role for the team workflow: talks to the human on a pull and on the team mod's next-action messages, single writer of the decisions file.

  • team-role-implementer

    Standing rules for the implementer role in the team workflow. Writes code in a worktree, runs fix rounds, creates MRs on its own decision. Preloaded into team-implementer.

  • team-role-investigator

    Standing rules for the investigator role in the team workflow. Read-only analysis, review, and design-document work. Preloaded into team-investigator.

  • team-role-tester

    Standing rules for the tester role in the team workflow. Diff review plus gate, live rounds, finding classification, manual-test partner. Preloaded into team-tester.

Slash commands (5)

  • /team:brief

    Compose a brief for an agent, fill its task section, send the kick-off, and note the status line in the plan file.

  • /team:init

    Start a team run for a ticket in this session: this session becomes the envoy, the orchestrator starts in a worker tab.

  • /team:release

    Clear and close finished agents, running the overlay release check first. Refuses to release a working agent.

  • /team:resurrect

    After a restart, relaunch the orchestrator and the team workers that came back without their team flags, resuming their sessions. Run it in the envoy session.

  • /team:status

    Read the roster, read idle agents that have not reported, and flag anything that needs attention. Run it in the envoy session.

Subagents (5)

  • team-envoy

    Talks to the human for a team run: presents the question queue on a pull and the next action on every run change, records decisions, relays requests to the orchestrator.

  • team-implementer

    Code-writing agent for the team workflow. Works in a worktree, runs fix rounds, creates MRs on its own decision. Started as a main session with --agent.

  • team-investigator

    Read-only analysis, review and design-document agent for the team workflow. Started as a main session with --agent.

  • team-orchestrator

    Runs a team: briefs agents, reads reports, asks the human through cards. Started by team-init in a worker tab.

  • team-tester

    Test and gate agent for the team workflow. Diff review, live rounds, finding classification, manual-test partner. Started as a main session with --agent.

Hooks (2)

  • SessionStart

    bash "${CLAUDE_PLUGIN_ROOT}/hooks/handlers/session-start.sh"

  • Stop

    bash "${CLAUDE_PLUGIN_ROOT}/hooks/handlers/stop-report.sh"

View source on GitHub

More from abrose/claude-plugins

All 5 plugins →

eddy

claude

Spike: the team workflow with card branches in the chat (fold) or in a pane, switchable. Runs beside the team plugin.

5 skills · 5 commands · 5 agents · hooks

0

Ships one output style combining Simplified Technical English (short, direct sentences) with an ADHD-friendly, action-first shape, and keeps the teaching depth. Select it from /config.

0

chebu-ui

claude

Personal Claude Code UI tweaks as a mod: bold numbered prompts, a /prompts pane, a separator after each turn, quiet tool rows, dim narration and framed answers.

0

Quota-spend projection engine for statuslines. Reads the Claude Code rate-limit payload, logs a rolling per-profile usage sample, and emits a JSON verdict per window (5h/7d): are you on pace to blow the limit before it resets, or coasting under it? Weekly projection counts active hours, not 24/7. Bring your own rendering.

0

Maintain team?

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 plugins5 plugins