AGI Super Team

Install a coordinated set of roles, skills, and outcome teams across local AI coding frameworks.

Stars
★ 121
Last updated
2d ago
License
MIT
Primary language
Python

At a glance

How it runs
CLIAgent plugin / skill
Works with
Portable with changesCodex · Claude Code (Partial support)
Cost
Free, no paid service needed
Setup effort
Low · running in minutes
You'll need
Node.js 18+npmShell / CLINetwork accessLocal filesystem
Typical use
A developer using Codex or Claude Code wants a product-delivery team that can produce a brief, implement a change, and review the result.
Not a fit if
  • Teams that need proof of runtime loading or task quality immediately after installation
  • Teams expecting the framework to publish, trade, or take irreversible actions automatically
Maintenance
Commits in the last 30 dayslast commit 2026-11-19latest release v1.8.0 (2026-11-04)
Source review
69/100 · Some gaps

What does this agent do, and when should you use it?

AGI Super Team is a versioned Agents + Skills team system for local AI agent frameworks, installed through a CLI into targets such as Codex, Claude Code, OpenClaw, and Hermes. It provides 14 top-level roles, 92 opt-in specialists, reusable Skills, and eight outcome-oriented team packs, with a CEO coordinator and an independent Governor reviewer. The installer previews destinations and conflicts by default and writes only after the user explicitly selects --install; adapters deliver native agents, skills, rules, or role packs according to each framework's capabilities. The --doctor command checks installed adapter artifacts, but installation does not prove that a client loaded them or that tasks perform well. The external framework supplies the model, credentials, tools, and execution, while publishing, trading, and other irreversible actions require human approval.

Users list targets with npx -y agi-super-team@latest --list-tools, then preview a specific adapter using --tool <id> to inspect files and conflicts; after review they run the same selection with --install, optionally add --connect for a connection receipt, and finish with --doctor. The default installs the 14 top-level roles and their assigned Skills; --with-subagents <executive> selects one executive's specialist group, while --all-subagents selects all 92 specialists. --skill <name> and --all-skills control Skills selection. Work starts from a user-defined outcome: the CEO sets acceptance checks and chooses a team, roles use Skills as methods and may delegate to specialists, the Governor independently reviews evidence, and the CEO synthesizes the result. If native multi-agent delegation is unavailable, the workflow should run sequentially and state that limitation.

  1. A developer using Codex or Claude Code wants a product-delivery team that can produce a brief, implement a change, and review the result.
  2. A content team has approved sources, an audience, and a channel, and needs to progress from an evidence brief to channel-ready drafts and claims review.
  3. A researcher has a consequential question and decision criteria and needs a research plan, evidence map, cited synthesis, and decision memo.
  4. A quantitative researcher wants to turn a hypothesis and historical data into a reproducible backtest specification, risk memo, and independent gate.
  5. An operations lead handling a bounded incident or delivery failure needs scoping, containment, verified recovery, and a post-incident review.

Key setup facts

Pulled from the install and usage notes below. The project README has the current values.

Install / run
  • npx -y agi-super-team@latest --list-tools
  • npx -y agi-super-team@latest --tool codex
  • npx -y agi-super-team@latest --tool codex --install --connect
  • npx -y agi-super-team@latest --tool codex --doctor
Runtime
Node.js 18+

How do you install or deploy this agent?

Requires Node.js 18+ and npm; the public npm commands fetch the package over the network. List targets and preview one explicit adapter, review its destination and conflicts, then install and connect it and run the diagnostic:

npx -y agi-super-team@latest --list-tools
npx -y agi-super-team@latest --tool codex
npx -y agi-super-team@latest --tool codex --install --connect
npx -y agi-super-team@latest --tool codex --doctor

How do you use this agent?

After installation, restart the target client or open a new task, then give it an outcome. This prompt asks the CEO to coordinate the team; the client determines whether work can run in parallel, and the plan should run sequentially when native delegation is unavailable:

Swarm agents: <your outcome>

Act as the CEO coordinator.
- Define the outcome and acceptance checks.
- Choose the smallest sufficient team.
- Parallelize only independent work with explicit file ownership.
- Use Skills as methods and require an independent Governor review.
- Synthesize one verified result.

If native multi-agent delegation is unavailable, execute the same role plan sequentially and say so. Do not publish, deploy, spend money, use credentials, or take irreversible action without my approval.

What are this agent's strengths and limitations?

Pros
  • One organizational contract targets multiple clients; the repository lists 22 adapters and maps delivery to each target's capabilities.
  • The 14 top-level roles, 92 opt-in specialists, and eight outcome teams let users start with a small roster and expand when evidence gaps justify it.
  • Preview is the default, --install is explicit, and the CLI provides conflict handling, managed-file backups, and a --doctor artifact check.
  • CEO coordination and independent Governor review are defined workflow roles, with human approval required for consequential actions.
Limitations
  • An adapter's presence does not establish runtime verification; the README says runtime evidence for primary adapters is pending, and --doctor does not test model behavior.
  • Frameworks have different capabilities and file layouts; some targets receive Agent-as-Skill or project rules, and Cursor and Antigravity are marked experimental.
  • The repository is not a model or runtime, so users still need the target framework's model, credentials, tools, and execution environment.
  • Nested Codex routing has depth and thread constraints; the README recommends max_depth = 2 and limits child tasks per wave.

How does this agent compare with similar options?

The repository also offers a curated Codex package and a legacy generic workspace materializer alongside the multi-client npm installer. Codex users can choose the package documented in .codex/INDEX.md; users who want inspectable role workspaces decoupled from a specific framework can preview and apply teams with the legacy install.sh path. These differ from the current npm adapter delivery path.

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
AGI Super Team This agent 69 · Some gaps CLIFree ★ 121 2d ago Python —
oh-my-agent 60 · Some gaps CLIFree + model costs ★ 1.5k 1d ago TypeScript Codex · Claude Code
Friday Agent Workbench 73 · Some gaps Desktop appFreemium ★ 83 today TypeScript Codex · Claude Code
Kungfu 64 · Some gaps CLIFree ★ 4.5k 2mo ago C++ Codex · Claude Code

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Some gaps
69/ 100 5-point scale 3.5 / 5
Trust 18/29
Reliability 8/14
Adaptability 15/18
Convention 14/18
Effectiveness 7/13
Verifiability 7/8
Why each dimension lost points
Trust18 / 29 · 3.1/5

The README describes preview-first installation, explicit install selection, managed-file boundaries, and backups as recovery aids rather than full snapshots. SECURITY.md calls for least-privilege credentials, credential protection, and human approval for external messages, transactions, deployments, and destructive actions. The redbook CLI sample can extract browser cookies or accept a manual cookie string; those sensitive-data paths are disclosed, but the sample does not show more detailed protections. npm and Python dependencies are present without dependency-audit or lockfile evidence in this packet, so dependency security scores lower. Upstream sources and hashes, licensing, and maintenance guidance have some evidence, but publisher identity remains unknown.

Reliability8 / 14 · 2.9/5

The README distinguishes file materialization from runtime verification and says runtime evidence for the five primary adapters is pending. It describes doctor checks, idempotence, and safer install targets, while the CLI sample includes error handling. However, the skills span varied runtime dependencies, endpoints, and environments; the packet also identifies a single-endpoint Binance failure and API latency issues. Dependency availability therefore cannot score highly.

Adaptability15 / 18 · 4.2/5

The README identifies eight outcome-oriented team kits, 14 core roles, optional specialists, and the installation form and maturity of 19 targets. It also documents delegation limits, runtime-verification status, and human approval for external effects. Sample skill triggers are understandable but uneven, and some examples depend on particular hosts, paths, regional services, or tools, limiting trigger precision and environment fit.

Convention14 / 18 · 3.9/5

The repository provides clear entry points for its architecture, adapters, skills, and installation guides. The README covers preview, installation, updates, doctor checks, and recovery, and discloses limitations and unverified status. The MIT license is explicit. Versioning uses multiple version schemes; the README intentionally avoids pinning the npm version, while sample skills have their own versions and histories, so naming and version consistency is imperfect. Contribution and maintenance guidance exists, but response times are not guaranteed and publisher identity is unknown.

Effectiveness7 / 13 · 2.7/5

Reusable roles, skills, outcome-oriented team kits, and a multi-framework installer provide practical structure. Adapter mappings and expected deliverables help make outputs usable. The value proposition has architectural support, but the 19 targets are not all runtime-verified, and some skills involve financial predictions, trading monitoring, or platform accounts with cookies. Benefits and maintenance costs therefore vary by use case; utility is moderate and cost-benefit scores lower.

Verifiability7 / 8 · 4.4/5

The README points to adapter mappings, source locks and hashes, validation tools, and evidence status, and distinguishes illustrative workflows from completed runs and structural installation from runtime verification. Plugin manifests, package.json, CI workflows, and the security policy provide some cross-source support. The packet is a sample, however, and many claims are not individually corroborated by corresponding implementation details or runtime receipts in the supplied material, so cross-source corroboration is not full.

Risks and how to mitigate them
  • Treat installation previews and structural checks as evidence of file deployment only; they do not establish that a target client loaded or correctly ran the content. Runtime verification for the primary adapters is explicitly pending.
  • Use skills involving prediction markets, trading, or platform accounts only with human review and an isolated environment; samples include a single-endpoint market-data dependency and browser-cookie access.
Evidence confidence: Low Reviewed Nov 11, 2026 Reviewed revision 8e7ac34a9b13 New commits since this review; the score may not cover them
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/agi-super-team.svg)](https://followagents.com/en/agents/agi-super-team)
<a href="https://followagents.com/en/agents/agi-super-team"><img src="https://followagents.com/badge/agi-super-team.svg" alt="FARS review"></a>

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

  1. v1.8.0 AGI Super Team 1.8.0
    Release notes excerpt
    What's Changed * feat(skills): 收录新媒体运营 10 项技能(Daniel IP + 对标拆解) by @aAAaqwq in https://github.com/aAAaqwq/AGI-Super-Team/pull/40 * feat: add agent-qa application testing skill by @pranshuchittora in https://github.com/aAAaqwq/AGI-Super-Team/pull/42 * Unbreak YouTube caption lookup in the bundled scraper skill by @johnisanerd in https://github.com/aAAaqwq/AGI-Super-Team/pull/41 * fix(ci): refresh Pages star statistics weekly instead of every 6 hours by @aAAaqwq in https://github.com/aAAaqwq/AGI-Super-Team/pull/39 …
    View on GitHub ↗
  2. v1.7.1 AGI Super Team v1.7.1
    Release notes excerpt
    npm 现在打包完整的 skills/ 目录 此前的 npm 包只把 team-manifest.json 指定少数技能打全,其余每个技能仅进 SKILL.md 入口——而 SKILL.md 正文里引用的 scripts/、references/、assets/ 却没进包,npm 用户按入口里的指令执行会找不到文件。本次把 package.json 的 files 里 172 条逐技能枚举换成单条 "skills/":每个技能都完整进包、开箱可用,不再依赖 manifest 成员资格,且不改动任何 SKILL.md(Gold Set 摘要与 catalog 树摘要保持不变)。决策与取舍见 ADR-0008。 - tarball 解压体积从约 13 MB 增至完整 skills/ 树(约 73 MB 落盘,压缩后更小)。.npmignore 已排除 __pycache__ / .venv / node_modules 等瞬时文件。 - 未来若受体积压力,应通过 .npmignore 排除大文件(如样例数据),而不是剥离指令引用的内容。 文档与技能线 …
    View on GitHub ↗
  3. v1.5.0 AGI Super Team v1.5.0
    Release notes excerpt
    250 commits since v1.4.2, including 7 new skills. New skills - muapi-video-gen — text-to-video through MuAPI's FLUX 3 endpoint. Preview-then-confirm by default: the request plan is printed and no network call happens unless --yes is passed, and the billable submission POST is never retried. Output URLs are HTTPS-validated (including after redirects) before download. - atlas-video-gen — video generation via Atlas Cloud. …
    View on GitHub ↗

FAQ

Does installation launch agents or verify task quality?
No. The installer materializes adapter files, and --doctor checks those artifacts; loading and task performance must still be verified in the target client.
Do I need a paid model API key?
The repository does not specify a model or API provider or state model costs. The selected external framework supplies model access and credentials.
Can I install every specialist?
Yes. Use --all-subagents --install with any --tool to install all 92 opt-in specialists, or use --with-subagents <executive> to select one executive's group.
What if my framework has no native multi-agent delegation?
The README says to execute the same role plan sequentially and disclose the limitation. Actual concurrency, nesting, and model access depend on the client.
Will it publish content or place trades automatically?
No. The workflow requires explicit human approval for publishing, trading, deployment, credential use, and irreversible actions.
Is AGI Super Team open source?
Yes. The source code is on GitHub (aAAaqwq/AGI-Super-Team) under the MIT licence.
Is AGI Super Team safe to use?
FollowAgents' FARS review (Nov 11, 2026) scored AGI Super Team 18/29 on trust and safety. None of the eight safety controls the review checks for was missing outright. The repository has changed since that review.
Is AGI Super Team still maintained?
Yes. The latest commit on GitHub is from 2026-11-19, within the last 30 days. The latest release is v1.8.0 (2026-11-04).
Who makes AGI Super Team?
AGI Super Team is published on GitHub by the account aAAaqwq. FollowAgents has not verified which person or company is behind that account.
Was this page helpful?
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related concepts