Release Immortal Memory v1.3.3

[HeiGeAi]

What changed

• Complete the reviewed Agent Context preview and compile lifecycle across the CLI, MCP, dashboard, and adapters.
• Replace retired judgment cache reads with current event-backed authority.
• Add an owner-only, redacted Feishu learning review reminder with explicit remote-write confirmation.
• Report the full pending review count and provide real Trust review routes.
• Let users bind confirmed and challenged memory references to real Context outcomes.
• Surface challenged references in the Trust ledger without automatically mutating Claim or Judgment authority.

Why

Immortal must show what an Agent actually consumed, what happened afterward, and which memories the result supports or challenges. The release keeps that feedback visible and auditable while preserving user confirmation as the only promotion path.

Validation

• Python 3.11 full suite: 1332 passed
• Focused product and Outcome regression: 211 passed
• Source privacy scan: private_scan=ok
• Wheel privacy scan: private_scan=ok
• Clean wheel/version governance: 2 passed
• Live installation reports immortal 1.3.3
• Live Trust API exposes challenged_memories and failed_outcome
• Browser Trust view renders the new review counter and category

Known boundaries

• Same-disk backup is not disaster recovery, so the local system remains in attention state.
• Complete owner, project, session, and Agent profile isolation requires a separate schema migration.
• Feishu interactive card callbacks require a trusted inbound callback service and are not included in this release.