This release fixes issues for SREs watching stability and regressions.
✓ No known CVEs patched in this version
Summary
AI summaryFixed false positive drift and compliance signals plus reduced usageGuidance verbosity.
Full changelog
Fixes
Drift false positive on completed tasks — The drift signal was firing even after a task finished, producing spurious checkpoint prompts as the conversation wound down. Now suppressed when checkpoint.currentStep starts with COMPLETE.
Compliance false positive — unused injection — injectedFactsAreTaskRelevant() was comparing injected facts against the stale session-level inferredTaskType instead of the current user message. Facts relevant to what was actually asked were incorrectly flagged as unused, incrementing consecutiveUnusedMemoryInjections and degrading compliance.
usageGuidance verbosity — The MANDATORY protocol reminder block was re-injected every turn even on fully healthy agents. Now suppressed when all compliance counters are zero (isComplianceHealthy), saving injection tokens on well-behaved hosts.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About nfemmanuel/iranti
Persistent shared memory for AI coding agents. Stores facts as `entity/key/value` triples with hybrid semantic search, task checkpoints, and conflict resolution — shared across Claude Code, Codex CLI, and GitHub Copilot.
Related context
Related tools
Beta — feedback welcome: [email protected]