What this repository is

AGI Super Team provides installable AI team packs, role instructions, and reusable skills for existing coding-agent harnesses.

It is not a model or an agent runtime. It provides versioned workspace configurations that help people plan, build, review, and communicate with explicit human approval gates.

Three constraints make the project different:

  • Inspect before write: the generic installer previews every destination and requires explicit --apply.
  • One source of truth: agents, kits, local requirements, and external recommendations come from a validated manifest.
  • Evidence before claims: compatibility and outcomes remain pending until a receipt matches the tested revision.

Trading, legal, security, medical, publishing, and deployment work still requires qualified human review.

Safe quick start

Clone a trusted revision and preview the Solo Founder pack. The installer requires Bash and Node.js; repository verification also requires npm and Python 3.

git clone --depth 1 --branch main https://github.com/aAAaqwq/AGI-Super-Team.git
cd AGI-Super-Team
./install.sh --source "$PWD" --destination /path/to/review-workspace solo-founder

Review the source, destination, selected agents, and planned files. Apply only when the preview matches your intent:

./install.sh --source "$PWD" --destination /path/to/review-workspace --apply solo-founder

The installer preserves existing persona files and skill directories. It rejects missing required skills, source symlinks, and destination symlinks before publishing staged files.

Avoid piping a remote script into a shell when you can inspect a pinned checkout first. See setup.md for prerequisites, updates, and recovery.

Preview → apply → verify

The animation is an illustrative storyboard with sanitized paths, not runtime evidence. Read the storyboard transcript, then use the receipt below for reproducible commands.

Five-minute installation receipt

The repository currently verifies a safe installation outcome. It does not yet claim that an installed team has produced a validated business result.

“Five-minute” describes the walkthrough scope, not a speed guarantee. This path evaluates the generic file installer; it does not configure a model or make a harness load generated workspaces automatically.

Create a disposable destination, prove preview wrote nothing, apply, then confirm the three expected role workspaces:

AGI_SOLO_DEST="$(mktemp -d "${TMPDIR:-/tmp}/agi-solo-founder.XXXXXX")"

./install.sh --source "$PWD" --destination "$AGI_SOLO_DEST" solo-founder
test -z "$(find "$AGI_SOLO_DEST" -mindepth 1 -print -quit)"

./install.sh --source "$PWD" --destination "$AGI_SOLO_DEST" --apply solo-founder
test -f "$AGI_SOLO_DEST/workspace-ceo/SOUL.md"
test -f "$AGI_SOLO_DEST/workspace-pe/SOUL.md"
test -f "$AGI_SOLO_DEST/workspace-cco/SOUL.md"

npm test
npm run validate -- --warnings-as-errors

Print and inspect $AGI_SOLO_DEST before moving or removing anything. A real customized destination requires its own backup and review; repeated apply is not an upgrade or rollback mechanism.

The installer stops after copying inspectable files. It does not launch agents, orchestrate roles, or create the outputs below. Use these only as optional manual prompts in a separately configured harness.

Workspace Responsibility Optional evaluation prompt
workspace-ceo Planning and quality gates Draft a decision memo with assumptions, alternatives, evidence gaps, and a human approval gate.
workspace-pe Engineering and delivery Propose a test-first implementation plan. Do not deploy or change production systems.
workspace-cco Launch communication Draft three launch-post variants with claim placeholders and a manual publishing checklist.

This receipt proves deterministic installation and repository integrity. Cross-harness task performance remains Validation pending until a public fixture is tied to the current main commit.

Explore

Start with a focused pack or guide instead of browsing the full skill inventory.

Path Use it for
Starter kits Small role combinations for a founder, content workflow, or quantitative research
Agents Persona, identity, workflow, and tool guidance for the generic installer
Practical guides Codex and Claude Code setup, compatibility, team choice, and workflow boundaries
Codex package A separately curated native package with its own manifest and sync policy
Cookbooks Longer learning material for content, prompts, research, and quantitative workflows
Skills The physical cross-harness library; use the validator rather than a hard-coded count

Starter kits

Kit Agents Intended use
Solo Founder CEO, PE, CCO Planning, engineering, and reviewed content drafts
Content Creator CCO, CDO, CMO Research, content drafts, and measurement plans
Quant Trader CQO, CDO, CFO Research, backtesting, and risk review—not live trading
full-team All manifest agents Broad evaluation; start smaller when possible

Choose a distribution

Surface Repository support Install path Evidence boundary
Generic/local workspace Supported by install.sh Preview, inspect, then apply Installer behavior is covered by integration tests
Codex Separate curated native package See .codex/INDEX.md Native package and generic kits are separate distributions
Claude Code Plugin manifest present Review .claude-plugin/ Confirm support in the installed client version
Cursor, Gemini, Kimi Metadata or manifests present Review the corresponding files Presence does not establish feature parity

Evidence and verification

Repository contracts calculate catalog facts from tracked files and the canonical manifest:

npm test
npm run validate
npm run validate -- --warnings-as-errors

A web page displays Verified only when its verification receipt matches the current main commit and every recorded check passes.

When making an outcome claim, link the reproducible input, fixture, revision, result, limitations, and rollback path. Tests are evidence of repository behavior, not a guarantee of business performance.

Repository architecture

AGI-Super-Team/
├── config/          # Canonical team manifest, schema, and removed-link provenance
├── agents/          # Role, identity, workflow, and tool guidance
├── skills/          # Tracked physical skills; symlinks are forbidden
├── starter-kits/    # Focused team selections
├── install.sh       # Preview-first generic workspace installer
├── scripts/         # Repository model, validator, and Pages data builder
├── tests/           # Repository, installer, site-data, and SEO contracts
├── docs/            # Evidence First Pages site and editorial guides
├── growth/          # Human-reviewed launch and measurement playbooks
└── assets/          # README, social preview, logo, and demo assets

The control flow is:

config/team-manifest.json
  → scripts/repository_model.py
  → install.sh + validator + tests

config/external-skill-sources.json
  → portable provenance for removed machine-local links

README text is not an inventory source. See the manifest, repository model, and contribution policy for the contracts.

Team topology

Founder / operator
└── CEO — coordination and quality gates
    ├── CTO / PE — architecture and implementation
    ├── CPO / CCO / CMO — product, content, and growth
    ├── CQO / CFO / CDO — quantitative research, finance, and data
    ├── CLO / CRO / CSO / COO — legal, research, sales, and operations
    └── Governor — independent review and escalation

Mentor names are creative framing. They do not imply affiliation, endorsement, or guaranteed imitation.

Safety boundaries

  • Never place credentials, private user data, browser sessions, or production configuration in a skill or issue.
  • Review third-party commands and dependencies before execution.
  • Keep financial workflows in research or paper-trading environments until independently validated.
  • Require explicit human approval for posts, messages, transactions, deployments, and destructive operations.
  • Report vulnerabilities privately through GitHub Security Advisories.

Project links

Star History

Star History