Skip to content

nfemmanuel/iranti

v0.3.22 Bugfix

This release fixes issues for SREs watching stability and regressions.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Summary

AI summary

Fixed compliance scorer false positives by ignoring task‑irrelevant injected facts.

Full changelog

v0.3.22

Fix: Compliance scorer false positive for task-irrelevant injected facts

When a session spans multiple task domains on the same entity (e.g., release work followed by marketing copy), the Attendant would surface entity-relevant but task-irrelevant facts (like CI status during a marketing turn). The compliance scorer then penalized the agent for not using those facts, producing false-positive compliance violations.

What changed:

  • Added task_irrelevant evidence kind to the memory attribution system
  • At injection time, the Attendant now captures inferredTaskType from the session brief
  • At scoring time, injectedFactsAreTaskRelevant() checks token overlap between the task context and injected fact keys/summaries
  • Facts identified as task-irrelevant count as "used" so the compliance counter does not increment
  • The compliance reason is set to injected_facts_not_relevant_to_current_task for observability

Testing:

  • New unit test suite test:compliance-task-relevance with 10 cases covering key-based relevance, summary-based relevance, edge cases, and mixed scenarios
  • Updated protocol enforcement tests (route + MCP) to use task-relevant fact data
  • Added to test:hardening-fast CI chain

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

Track nfemmanuel/iranti

Get notified when new releases ship.

Sign up free

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.

All releases →

Beta — feedback welcome: [email protected]