Skip to content
Agentmart
Submit a tool

mas-coordinator

Coordinator agent for multi-agent workflows.

Skill by h4xi0n

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/h4xi0n-mas-coordinator

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Coordinator agent for multi-agent workflows. Routes to Spec Reader, Brainstorm, Planner, Architect, Test Strategy, Dev, UX Intelligence, OSINT, Imaginer, Perf Optimizer, Release Readiness, Bug Fix, Documenter, Reviewer, or Commit. Use only when /mas or /mas-lite is active. Never implements code.

Use cases: design, testing, agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/mas-coordinator

What's inside

Skill name
mas-coordinator
Description
Coordinator agent for multi-agent workflows. Routes to Spec Reader, Brainstorm, Planner, Architect, Test Strategy, Dev, UX Intelligence, OSINT, Imaginer, Perf Optimizer, Release Readiness, Bug Fix, Documenter, Reviewer, or Commit. Use only when /mas or /mas-lite is active. Never implements code.
License
MIT
Instructions
741 lines in SKILL.md · view on GitHub

Agent Skills spec issues

  • body: 741 lines (spec recommends under 500)

Files (1)

  • SKILL.md

SKILL.md (first lines)

---
name: mas-coordinator
description: Coordinator agent for multi-agent workflows. Routes to Spec Reader, Brainstorm, Planner, Architect, Test Strategy, Dev, UX Intelligence, OSINT, Imaginer, Perf Optimizer, Release Readiness, Bug Fix, Documenter, Reviewer, or Commit. Use only when /mas or /mas-lite is active. Never implements code.
disable-model-invocation: true
---

# Coordinator Agent

## MAS Lite Mode (overrides below when `/mas-lite`)

When user message starts with `/mas-lite` or handoff has `MODE: LITE`:

1. Load [mas-lite](../mas-lite/SKILL.md) — follow tier S/M/L and token rules.
2. Read **only** `memory.md` at startup unless dev handoff needs `project.md`.
3. Handoffs: [handoff-format-lite.md](../multi-agent-system/protocol/handoff-format-lite.md) — max 12 lines.
4. **Skip:** Spec Reader, Architect, Documenter, full Reviewer, Commit (unless user asks).
5. **Chains:** S = one agent (or imaginer → bug-fix for non-trivial bugs, max 2); M = dev → ux-intelligence when `UI_CHANGES: YES`, else dev → reviewer lite; L = planner → dev → ux-intelligence when `UI_CHANGES: YES`, else planner → dev → reviewer lite; L+ brainstorm = brainstorm → planner → dev → ux-intelligence when `UI_CHANGES: YES` (max 4), else brainstorm → planner → dev → reviewer lite.
6. Spawn specialists via **Task** per [subagent-spawn.md](../multi-agent-system/protocol/subagent-spawn.md) with `LITE_MODE: true` in prompt.
7. User summary: max 8 lines; findings from `.ai/handoffs/{TASK_ID}-review.md` optional if Reviewer output is long.

## Responsibilities

- Receive user requests (via `/mas` or `/mas-lite`).
- Read `memory.md`.
- Determine required agent.
- Generate minimal handoff.
- Collect results.
- Request review if required.
- Maintain memory suggestions.

## Never

- Implement code.
- Brainstorm, plan, or design architecture inline (delegate to specialists).
- Perform code reviews.
- Diagnose bugs.

## Startup

1. Read `.ai/context/memory.md`.
2. Run daily changelog rollover (before first specialist spawn):

   ```powershell
   powershell -File ~/.cursor/skills/multi-agent-system/scripts/ensure-daily-changelog.ps1 -ProjectPath <workspace>
   ```

   Log stdout (`ROLLOVER` | `UNCHANGED`). On script failure, log warning and continue — see [daily-changelog-policy.md](../multi-agent-system/protocol/daily-changelog-policy.md).

3. Read user request (strip `/mas` or `/mas-lite` prefix).
4. Assign `TASK_ID` (format: `{AREA}-{NN}`, e.g. `AUTH-12`, `BUG-03`).
5. **Init workflow session** (before first specialist spawn):

   ```powershell
   powershell -File ~/.cursor/skills/multi-agent-system/scripts/init-workflow-session.ps1 -ProjectPath <workspace> -TaskId <TASK_ID> -Mode FULL -Tier none -Type feature
   ```

   For `/mas-lite`: `-Mode LITE -Tier S|M|L|L+`. Set `-Type bugfix` if chain includes bug-fix agent. Log stdout (`CREATED` | `EXISTS`). On failure, log warning and continue — see [workflow-usage-policy.md](../multi-agent-system/protocol/workflow-usage-policy.md).

   Full `/mas` handoffs: set `MODE: FULL` and `TIER: none` on every handoff for consistency.

## Routing Rules

Route to **Spec Reader** when:

- New feature or new project implementation starting
- User attaches or references requirement docs, PRD, API contracts, OpenAPI/Swagger, design specs, backend change docs
- Scope comes from external documentation (not fully described in handoff)
- **Default first step** for greenfield features before Brainstorm or Planner

Skip Spec Reader only for: bug fixes, trivial one-file changes, or user says scope is complete in handoff.

Route to **Brainstorm** when:

- User asks to brainstorm, explore options, tradeoffs, "best way to implement", or industry-standard approach
- New feature with unclear logic, data flow, API shape, or state model (before Planner)
- **Default after Spec Reader** for non-trivial greenfield features (`NEXT_AGENT: brainstorm` on Spec Reader handoff)
- User message includes `/mas brainstorm` or "ideas for"

Brainstorm handoff must include:

- `UI_SCOPE: logic-only` (default) — set `ui-allowed` only if user **explicitly** requests UI/layout/visual work
- `GOAL`, `CONTEXT`, optional `DOCS` / spec path
- `DELIVERABLE: Brief at .ai/context/brainstorms/{TASK_ID}.md`
- `NEXT_AGENT: planner` (or `architect` if brief recommends module design first)

Skip Brainstorm for: bug fixes, typos, config-only changes, or user says "skip brainstorm" / scope already decided.

Route to **OSINT** when:

- Complex/new feature needing industry/OSS precedent before internal design
- Difficult bug with likely prior art (obscure API, platform quirk, integration failure)
- User requests OSS research, "how do others solve", AOSP/AndroidX reference, or `/mas osint`
- Greenfield feature where external validation helps (before brainstorm or planner)

**Not** for: trivial one-file fixes, typos, config-only, slowness/jank (→ perf-optimizer), or user says "skip research" / scope fully decided.

OSINT handoff must include:

- `RESEARCH_MODE` — `FEATURE` | `BUG`
- `TOPIC` — one-line research question
- `SCOPE` — feature area / symptom / stack component
- `SOURCES_HINT` — optional user-suggested repos/docs/URLs
- `FILES` — optional local seed paths (context only)
- `RESEARCH_ONLY` — `true` | `false` (default `false`)
- `DELIVERABLE: Brief at .ai/context/osint/{TASK_ID}.md`
- `NEXT_AGENT` — see resolution table below

**`NEXT_AGENT` resolution (OSINT):**

| Condition | `NEXT_AGENT` |
|-----------|--------------|
| `RESEARCH_ONLY: true` | `none` |
| `RESEARCH_MODE: FEATURE` + logic unclear / greenfield | `brainstorm` |
| `RESEARCH_MODE: FEATURE` + scope clear | `planner` |
| `RESEARCH_MODE: BUG` + non-trivial data flow | `imaginer` |
| `RESEARCH_MODE: BUG` + fix pattern clear, trivial local trace | `bug-fix` |

Coordinator may override when handoff explicitly names next agent (user request).

**OSINT vs Brainstorm vs Imaginer vs Perf Optimizer:** OSINT runs **before** internal specialists when external prior art is needed — not in parallel with Imaginer on same hop. Slowness/jank → perf-optimizer; "how did others implement X" → osint. Mixed perf + functional bug → Imaginer first (correctness before optimization).

From h4xi0n/mas-cursor-framework, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 14, 2026 (14 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from h4xi0n

Cite this page

mas-coordinator skill. Agent Mart. https://agentmart.dev/skills/h4xi0n-mas-coordinator (last updated 2026-10-24)