Skip to content

v0.1 — Governance Harness Baseline

Choose a tag to compare

View all tags
@falltwo falltwo released this 20 Jul 22:30
v0.1
a096962

v0.1 records the first cleaned public snapshot of the AI Agent ERP governance harness. It establishes the control and evidence path used by later supply-chain workflows.

Included capabilities

  • Tool registration, per-Agent allowlists, role checks, and risk classification through the Tool Gateway
  • Approval handling for protected writes, with explicit pending and denied states
  • Audit records with hash-chain integrity checks
  • Transaction boundaries for multi-step SQLite writes
  • Web and LINE access controls with restricted tool exposure
  • Model fallback, usage logging, prompt-injection safeguards, and error masking

Verification

  • Public snapshot commit: a0969624425eeeebaa263e9724219d65a033fe16
  • Local verification on 2026-07-21: 56 passed
  • CI workflow included in the public snapshot

Scope and limitations

  • This release focuses on the governance harness. Supply-chain intelligence, proposals, approval evidence, and ERP execution were not yet presented as the complete L1->L2->L3 workflow introduced in v1.0.
  • The public tag points to the first cleaned public snapshot. It is not a byte-for-byte publication of earlier internal development history.
  • The deployment boundary is a local proof of concept based on SQLite; production IAM, shared-database multi-tenancy, and distributed execution are outside this release.

Comparison with v1.0

Area v0.1 v1.0
Primary scope Governance harness Governed supply-chain decision workflow
User flow Role- and tool-level controls L1 observation -> L2 proposal -> L3 approval and execution
Approval model Generic protected writes Domain Proposal, Approval, Execution, and receipt chain
Public verification 56 passing tests 327 passing tests
Assets 2
Loading