Your AI agent is smart. But it still:

• ❌ Starts coding before understanding the goal
• ❌ Wastes tokens on steps that don't matter
• ❌ Gives you outputs you can't replicate or audit
• ❌ Has no research phase before execution
• ❌ Produces deliverables that look like AI slop

StratOS fixes this at the root. It's not a prompt. It's an operating system — a meta-layer that governs how your agent thinks before it acts.

StratOS intercepts every request and runs a 4-phase algorithm before any output is produced:

Phase 0  →  Intercept & Classify
             Parse the Goal Sentence. Apply the Garbage Filter.

Phase 1  →  The Algorithm (Question → Delete → Simplify → Accelerate → Automate)
             Eliminate waste before executing.

Phase 2  →  Research First (No Exceptions)
             Nothing is built until research is done.

Phase 3  →  Execute
             Heads-down on the goal. Design and production together.

Phase 4  →  Mandatory Output Package
             Action log + Research summary + Cost analysis + PRD.

Every task ends with a complete, auditable, replicable output package.

bash <(curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/scripts/install.sh)

mkdir -p ~/.claude/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.claude/skills/stratos/SKILL.md

# Global install (available across all your OpenClaw agents)
mkdir -p ~/.openclaw/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.openclaw/skills/stratos/SKILL.md

# Or: project-level (for a specific agent workspace)
mkdir -p .agents/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o .agents/skills/stratos/SKILL.md

Or search stratos on ClawHub and install in one click.

mkdir -p ~/.cursor/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.cursor/skills/stratos/SKILL.md

mkdir -p ~/.gemini/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.gemini/skills/stratos/SKILL.md

mkdir -p ~/.codex/skills/stratos && \
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.codex/skills/stratos/SKILL.md

OpenClaw uses a skills system where skills are stored as directories containing a SKILL.md file — which is exactly StratOS's format. StratOS is a drop-in skill for OpenClaw.

OpenClaw is a powerful personal AI assistant that runs across WhatsApp, Telegram, Discord, Slack, and more. The problem: it's powerful, but process-unstructured by default. StratOS adds the missing execution layer.

With StratOS installed in your OpenClaw agent:

• Every request gets a Goal Sentence before execution
• The Garbage Filter eliminates wasted actions before they run
• Research runs first — no more hallucinated plans
• Every completed task includes a full PRD

Option 1: ClawHub (recommended)

Search "stratos" on ClawHub → Install

Option 2: Manual global install

mkdir -p ~/.openclaw/skills/stratos
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.openclaw/skills/stratos/SKILL.md

Option 3: Per-workspace (for a specific agent)

mkdir -p .agents/skills/stratos
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o .agents/skills/stratos/SKILL.md

After installing, send /status to your OpenClaw agent and it will list StratOS as an active skill.

nanobot is a lightweight OpenClaw-compatible agent. StratOS works with it:

mkdir -p ~/.nanobot/skills/stratos
curl -sL https://raw.githubusercontent.com/Techiral/stratos/main/SKILL.md \
  -o ~/.nanobot/skills/stratos/SKILL.md

StratOS is a meta-skill — a SKILL.md file that the agent reads before any other skill, tool, or action.

Any step is eliminated if it is:

• Not repeatable
• Replaceable by something cheaper or faster
• Not definable in one clear sentence
• Not solving the stated problem
• Using more tokens than needed
• Scheduled beyond 1 week without decomposition

Before executing anything, StratOS forces a single sentence:

"The goal is to [verb] [object] so that [measurable outcome]."

If this sentence can't be written, execution stops.

stratos/
├── SKILL.md              # The OS itself
├── README.md             # This file
├── CONTRIBUTING.md       # How to contribute
├── LICENSE               # MIT
├── llms.txt              # AI-readable index of this repo
├── CITATION.cff          # Academic citation info
├── scripts/
│   ├── install.sh        # Universal installer
│   └── validate.sh       # SKILL.md validator
├── docs/
│   ├── INSTALL.md        # Per-agent install guide (9 agents)
│   ├── HOW_IT_WORKS.md   # Architecture deep dive
│   └── PHILOSOPHY.md     # SpaceX + Machiavelli reasoning
└── .github/
    ├── workflows/validate.yml
    ├── FUNDING.yml
    └── ISSUE_TEMPLATE/

StratOS is built on one insight: LLMs are brilliant reasoners trapped in terrible processes.

Inspired by SpaceX's engineering review culture (question every requirement, delete before adding) and Machiavellian execution theory (stability > everything, fortuna & virtù).

Read the full philosophy: docs/PHILOSOPHY.md

PRs welcome. See CONTRIBUTING.md.

Want to add support for a new agent? Open an issue — I'll add it within 24 hours.

MIT — use it, fork it, build on it.