v1.110.0

What's Changed

Added

• ll-loop audit-meta subcommand — Detects evaluator-trivial failure mode in meta-loops; includes updated loop-specialist agent and docs. (ENH-1700)
• ll-action list --json exposes args hint — Per-skill input schema now surfaced for agent callers. (ENH-1660)
• Meta-eval telemetry write and archive — Meta-loop evaluation results written to a persistent telemetry store and archived per run. (ENH-1699)
• --show-diagrams options wired into ll-loop show — Topology, preset, and modifier flags available in the show subcommand.
• Learning Tests Guide — New documentation for the learning test registry added to docs.

Fixed

• Missing default_timeout in FSM loop YAMLs — Added default and per-state timeouts to unprotected loops to prevent indefinite hangs. (BUG-1724)
• hitl-md loop tests reflect finalize state — Corrected test drift caused by FSM shape change.

Changed

• hitl-md inline highlights + popover affordances — Diff highlights rendered inline with popover UI; final HTML copied to cwd on completion.
• Prompt inputs truncated to single-line preview in default ll-loop output for cleaner display.

---

Full Changelog: v1.109.0...v1.110.0