AGI Agent OS

Governance-first template stack for controlled AI agent workspaces.

中文说明请见: README.zh-CN.md

Project Philosophy

AGI Agent OS is built on one core idea: agent capability must be paired with explicit governance.

The stack treats a workspace as an operational system, not just a prompt folder. It defines clear authority boundaries, policy gates, mutation controls, and accountability artifacts so teams can run agents in a controllable way.

Core principles:

• explicit authority before action
• policy-first execution
• controlled mutation for high-trust files
• append-oriented audit discipline
• typed memory with provenance and staleness checks
• validator-friendly contracts over ambiguous prose

What This Repository Contains

• Bilingual specs (English + Chinese)
• Versioned release artifacts
• Controlled OpenClaw example pack
• Pass/fail fixtures for validator regression
• Commercial usage guidance and legal baseline files

Why It Exists

Most agent projects fail operationally at three points:

• policy is implicit and inconsistently enforced
• protected changes happen without robust review/traceability
• runtime validation is ad hoc and difficult to automate

This repository addresses those gaps by combining specification, fixtures, and operational release discipline.

Versioning & Evolution

The public baseline starts at v1.0.0, derived from iterative local development history (v0.0.1 -> v0.1.2).

Evolution focus across those iterations:

• governance model foundations
• activation-gate semantics
• validator matrices and finding families
• OpenClaw controlled workspace templates
• runnable + invalid fixtures for CI assertions

Entry points:

• Specs index: INDEX.md
• Current spec (EN): docs/specs/AGI-Agent-OS-Template-Stack-v1.0.0.en.md
• Current spec (ZH): docs/specs/AGI-Agent-OS-Template-Stack-v1.0.0.zh-CN.md
• Release notes: docs/RELEASE-NOTES.md
• Changelog: CHANGELOG-v1.0.0.md

OpenClaw Quick Start

1. Copy examples/v1.0.0-openclaw-controlled/workspace-template/ into your OpenClaw workspace.
2. Confirm governance core files exist (POLICY.md, CHANGE-CONTROL.md, STACK-MANIFEST.yaml, AUDIT/index.md, MEMORY/index.md).
3. Validate manifest consistency before protected activation.
4. Use workspace-template-invalid/ with VALIDATION-EXPECTATIONS.md to test failure-path behavior.

Commercial Use Notes

For business use, review these files first:

• docs/COMMERCIAL-GUIDELINES.md
• docs/COPYRIGHT.md
• NOTICE
• LICENSE
• docs/REFERENCES.md

Recommended production controls:

• protected branch and mandatory PR review
• CI validator checks before merge/release
• versioned release tags (vX.Y.Z)
• immutable audit-chain conventions in deployed workspaces

References & Citation

External interoperability references are tracked in docs/REFERENCES.md.

Citation policy:

• keep references attributable
• avoid large verbatim copy from third-party docs
• refresh reference links at each significant release

Privacy Baseline

This public repository excludes local machine paths and local environment identifiers.

License

Licensed under Apache License 2.0. See LICENSE.