Tags: mnemom/aip
Tags
feat(governance): add triggering_governance_signal_id to IntegritySig… …nal (ADR-048 PR9) (#55) Optional cross-link to a Mnemom platform governance_signals row that contextualizes an integrity result. Set by the host platform when, for example, a sideband.coherence drop preceded a boundary_violation — suggesting the violation reflects team-level drift rather than an agent-isolated lapse. Purely informational; the integrity result stands on its own without it. Surfaced in audit pipelines and operator dashboards (PR10 UI). Format: gs-{12-hex} matching governance_signals.id (mig 189–192). Bumps @mnemom/agent-integrity-protocol to 1.1.0 (non-breaking minor). verdict / proceed / recommended_action semantics unchanged. Co-authored-by: alexgarden-mnemom <support@mnemom.ai> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Improve analyzer quality: chain-of-thought, false positive fix, cache… … optimization - Add chain-of-thought instruction to system prompt (A1) - Fix false positive "API keys" pattern with disambiguation principle (A2) - Move evaluation instructions from user message to system prompt for better prompt cache hit rate (A3) - Add JSON sanitization for trailing commas in checkIntegrity (A4) - Add calibration example for clear verdicts (A5) - Bump both packages to v0.2.1 All changes applied to both TypeScript and Python SDKs in parity. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
PreviousNext