Skip to content
Back to search

agent-team-kit

MarketplaceWorks with Claude CodeGitHub

A four-agent team and a side pane for Claude Code.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 21 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/Dalibex/agent-team-kit with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/Dalibex/agent-team-kit'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/dalibex--agent-team-kit/badge.svg)](/listings/dalibex--agent-team-kit)

HTML

<a href="/listings/dalibex--agent-team-kit"><img src="/api/listings/dalibex--agent-team-kit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

In Claude Code (the CLI, or the Code tab of the desktop app), type:

/plugin marketplace add Dalibex/agent-team-kit
/plugin install agent-team@agent-team-kit

Open a new session, or run `/reload-plugins`. The pane appears beside the conversation with the team asleep. Then add the team rules to your `CLAUDE.md` so Claude delegates to the team:

/agent-team rules

Try it: ask Claude to "use the explorer to list the files in this folder" and watch it wake up.

Requirements: Claude Code with plugin support. Optional, for the GitHub tab: [GitHub CLI](https://cli.github.com/) signed in (`gh auth login`); without it the tab still shows the local git state.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 14 lines
  1. {
  2. "name": "agent-team-kit",
  3. "description": "A four-agent team and a side pane for Claude Code.",
  4. "owner": {
  5. "name": "Dalibex"
  6. },
  7. "plugins": [
  8. {
  9. "name": "agent-team",
  10. "source": "./plugin/agent-team",
  11. "description": "Four team agents (explorer, implementer, tester, reviewer) with project context in every brief, and a side pane in tabs: the team at work, usage limits and token spend, skill packs, builds and GitHub state. English and Spanish."
  12. }
  13. ]
  14. }
README.md · README · 203 lines
  1. # agent-team-kit
  2. [![CI](https://github.com/Dalibex/agent-team-kit/actions/workflows/ci.yml/badge.svg)](https://github.com/Dalibex/agent-team-kit/actions/workflows/ci.yml)
  3. [![Release](https://img.shields.io/github/v/release/Dalibex/agent-team-kit)](https://github.com/Dalibex/agent-team-kit/releases/latest)
  4. [![License: MIT](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
  5. A four-agent team and a side pane for [Claude Code](https://claude.com/claude-code).
  6. The plugin ships the agents, gives every brief the context it needs, and shows
  7. the team at work beside the conversation, with your usage limits, your skill
  8. packs, your builds and your repository's state. In English or Spanish.
  9. ## Contents
  10. - [Install](#install)
  11. - [The team](#the-team)
  12. - [The pane](#the-pane)
  13. - [Skill packs](#skill-packs)
  14. - [Team rules](#team-rules)
  15. - [Settings](#settings)
  16. - [Update and uninstall](#update-and-uninstall)
  17. - [Troubleshooting](#troubleshooting)
  18. - [Development](#development)
  19. - [License](#license)
  20. ## Install
  21. In Claude Code (the CLI, or the Code tab of the desktop app), type:
  22. ```
  23. /plugin marketplace add Dalibex/agent-team-kit
  24. /plugin install agent-team@agent-team-kit
  25. ```
  26. Open a new session, or run `/reload-plugins`. The pane appears beside the
  27. conversation with the team asleep. Then add the team rules to your
  28. `CLAUDE.md` so Claude delegates to the team:
  29. ```
  30. /agent-team rules
  31. ```
  32. Try it: ask Claude to "use the explorer to list the files in this folder" and
  33. watch it wake up.
  34. Requirements: Claude Code with plugin support. Optional, for the GitHub tab:
  35. [GitHub CLI](https://cli.github.com/) signed in (`gh auth login`); without it
  36. the tab still shows the local git state.
  37. ## The team
  38. Four agents ship inside the plugin, and Claude sees them as
  39. `agent-team:<name>`:
  40. | Agent | Model | Does |
  41. | --- | --- | --- |
  42. | `explorer` | Haiku | Searches and reads code, never edits. Sent instead of Claude reading many files. |
  43. | `implementer` | Sonnet | Makes an agreed change and verifies it with tests or a build. |
  44. | `tester` | Sonnet | Writes and runs tests; reports bugs in production code instead of fixing them. |
  45. | `reviewer` | Opus | Reviews a finished diff before a commit; the costliest, so only for non-trivial changes. |
  46. - **Context in every brief.** The plugin appends the project folder, its
  47. branch and the skills you picked in the pane; each agent loads only the
  48. ones its task needs.
  49. - **Questions go up, not sideways.** An agent that lacks something stops and
  50. answers `QUESTIONS:`. Claude continues that same agent with SendMessage, so
  51. it keeps what it already read, or asks you when the decision is yours.
  52. - **Delegation.** Out of the box, Claude Code's Agent tool tells Claude not to
  53. delegate unless asked. The plugin replaces that paragraph with one that
  54. presents the team and points to your rules. The Delegate switch in the Team
  55. tab turns it off, from the next conversation on.
  56. ## The pane
  57. The drawing shows Claude and its agents as pixel sprites: an agent sleeps
  58. until it gets a task, works while it runs and finishes with confetti and a
  59. sound of its own. Beside it, your usage limits: the 5-hour window with the
  60. time it fills at the current pace, the weekly window and the session's
  61. context, amber at 80 % and red at 95 %, with details on hover and a button to
  62. compact the context.
  63. Under the drawing, one tab at a time; a hidden tab's button flags what is
  64. happening in it, and the open tab is remembered.
  65. - **Team**: tokens spent per role this session and the team's share; one row
  66. per agent with its task, time, model and the skills it loaded; its brief,
  67. tokens and report when unfolded. Rows also flag waste: a brief without Goal
  68. or Why, a repeated brief that could have continued an agent, an Opus review
  69. of a change of ten lines or fewer, and Claude reading more than five files
  70. without the explorer.
  71. - **Skills**: your packs and loose skills (below).
  72. - **Jobs**: installs, builds, tests and downloads with their time and, in the
  73. background, their real progress; a system notification when something long
  74. finishes.
  75. - **GitHub**: the branch, uncommitted changes, the PR with its checks and a
  76. link to the failing one, PRs waiting for your review, and links to the
  77. repository; a notice when the CI it was watching finishes.
  78. `/agent-team` hides or shows the pane. In a terminal the pane is drawn as
  79. text. The pane calls no model and costs no tokens.
  80. ## Skill packs
  81. Instead of typing `/skill` before each request, pick skills in the Skills
  82. tab:
  83. - **Packs**: twelve packs, each with a name you type and up to 24 skills.
  84. When a pack is picked, every request you send offers its skills to Claude,
  85. which loads only the ones the request needs.
  86. - **Loose skills**: eight slots behind the "Loose skills" button; a picked
  87. one is loaded whole with every request.
  88. Picks are remembered per project folder; packs and slots are the same
  89. everywhere. A skill that is no longer installed is flagged and not sent. The
  90. Team tab shows which skills each agent actually loaded, so you can tell
  91. whether a pack is too broad.
  92. ## Team rules
  93. `/agent-team rules` adds the team rules to your global `CLAUDE.md`
  94. (`~/.claude/CLAUDE.md`, or the folder in `CLAUDE_CONFIG_DIR`), between
  95. `<!-- agent-team:reglas:inicio -->` and `<!-- agent-team:reglas:fin -->`,
  96. leaving the rest of the file as it was and keeping a copy of it first
  97. (`CLAUDE.md.agent-team-<time>.bak`, the last five). Run it again after an
  98. update to bring them up to date; `/agent-team rules off` removes them. It
  99. changes nothing, and tells you why, when the file is not UTF-8, when the
  100. markers are broken, or when it finds an older team section without markers
  101. (remove that one first). Type it yourself: it does not run from another
  102. plugin or a message. The rules tell Claude:
  103. - when to use each agent (the explorer past five files, the implementer for
  104. agreed changes beyond a couple of edits, the tester after non-trivial
  105. changes, the reviewer before a non-trivial commit);
  106. - the brief template: Goal, Why, Files, Done when, Limits;
  107. - to answer `QUESTIONS:` by continuing the same agent, and to check what each
  108. agent returns before accepting it.
  109. ## Settings
  110. `/plugin configure agent-team@agent-team-kit`, or the plugin's settings in
  111. the desktop app:
  112. | Option | Values | Default |
  113. | --- | --- | --- |
  114. | `teamName` | Any name: "Ana" titles the pane "ANA TEAM" | empty: "AGENT TEAM" |
  115. | `language` | `auto`, `en`, `es` | `auto`: Spanish when your system is in Spanish, English otherwise |
  116. | `updateCheck` | `true`, `false` | `true`: once a day, ask GitHub for the latest release |
  117. | `sessionLog` | `true`, `false` | `true`: when a session that used the team ends, append a summary of its agents to `agent-team/runs.md` in the Claude config folder |
  118. ## Update and uninstall
  119. The pane tells you when a new version is out. To update, in a terminal:
  120. ```bash
  121. claude plugin marketplace update agent-team-kit
  122. claude plugin update agent-team@agent-team-kit
  123. ```
  124. then run `/reload-plugins` in Claude Code, or open a new session. You can also
  125. enable auto-update for the `agent-team-kit` marketplace in `/plugin` →
  126. Marketplaces. To uninstall, run `/agent-team rules off` first if you added
  127. the rules, then:
  128. ```
  129. /plugin uninstall agent-team@agent-team-kit
  130. /plugin marketplace remove agent-team-kit
  131. ```
  132. ## Troubleshooting
  133. - **No pane**: open a new session; `/agent-team` shows it if it was hidden.
  134. - **Two panes**: another copy of agent-team is enabled; keep one in
  135. `/plugin`.
  136. - **The GitHub tab says `gh` does not answer**: install GitHub CLI and run
  137. `gh auth login`.
  138. - **Claude does not delegate**: check that the rules are in your `CLAUDE.md`
  139. (`/agent-team rules`) and that the Delegate switch in the Team tab is on.
  140. ## Development
  141. ```text
  142. .claude-plugin/marketplace.json the marketplace /plugin reads
  143. plugin/agent-team/ the plugin
  144. hooks/ register.tsx (events), pane.tsx (UI), state.ts (pure logic),
  145. actions.ts (side effects), i18n.ts (English and Spanish),
  146. drawing.ts, look.ts, usage.ts, jobs.ts, github.ts
  147. agents/ the four agents
  148. tests/ the test suite
  149. ```
  150. ```bash
  151. claude plugin test plugin/agent-team
  152. claude plugin validate plugin/agent-team
  153. claude plugin validate .
  154. ```
  155. The plugin's own [README](plugin/agent-team/README.md) explains its design
  156. and the engine rules every module must follow. See
  157. [CONTRIBUTING.md](CONTRIBUTING.md) for branches, commits, pull requests and
  158. releases, and [SECURITY.md](SECURITY.md) to report a vulnerability.
  159. ## License
  160. [MIT](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.