| v0.5.0 |
2026-08 |
- Expanded community threat list to 45 signals + 8 compact patterns (up from 25 + 2).
- New obfuscation decoders for HTML-entity and unicode-escape hiding.
- Added optional code-change review gate (
POST /api/v1/code-change) — scan an AI-generated code diff and record a signed, tamper-evident evidence verdict.
- 32 new threat-expansion smoke tests + 5 code-change gate tests.
- Full suite now 116 passing (was 78).
|
| v0.4.0 |
2026-08 |
- Rebranded to AgentInterdict.
- Repositioned as runtime enforcement for autonomous AI agents.
- Introduced the INGEST → TRUST → MEMORY → AUTHORITY → INTERDICT → EXECUTE pipeline.
- CLI renamed to
interdict; package/module renamed to agentinterdict.
|
| v0.3.0 |
2026-05 |
- Action-time revalidation of authority.
- Credential-not-memory enforcement (reject or redact before persistence).
- Quarantine + atomic containment of suspicious memories.
- Fail-closed tamper detection — runtime flips to
lockdown on tamper.
- Operator / runtime privilege separation.
|
| v0.2.0 |
2026-03 |
- Origin-bound authority + no derivation amplification.
- Guarded retrieval (retrieval ≠ permission).
- Trust graph.
- REST API:
scan, memories, search, action-check, stats.
- Hermes / OpenClaw / MCP integrations.
|
| v0.1.0 |
2025-12 |
- Initial release.
- Local memory gateway.
- Static transparent risk rules.
- Single-operator GUI.
|