v1.114.0

What's New in v1.114.0

Added

• generator-evaluator oracle sub-loop — Drives a generate → evaluate → revise cycle for any skill; converts 5 harness loops to thin wrappers. Includes playwright_screenshot fragment. (ENH-1775)
• enumerate-and-prove oracle — Enumerates candidates and proves each satisfies a predicate; converts 2 integration loops. (ENH-1776)
• convergence_gate + ll_rubric_score fragments — Shared fragments for convergence exit predicates and numeric rubric scoring; converts 5 convergence callers. (ENH-1776)
• implement-issue-chain oracle — Reusable chained-implementation sub-loop extracted from auto and sprint-refine-and-implement.
• research-coverage oracle — Coverage-driven research sub-loop extracted from deep-research loops.
• diff_stall_gate fragment — Extracted evaluator config fragment for diff stall detection.
• queue_pop + queue_track fragments — lib/common.yaml fragments for queue-based issue processing. (ENH-1875)
• /ll:scope-epic skill — Decomposes a theme or product area into a structured EPIC with scaffolded child issues.
• /ll:review-epic skill — Audits EPIC health: completeness, dependency ordering, and implementation risk.
• ll-workflows propose subcommand — Step 3 of the workflow analysis pipeline; proposes automation loop designs from identified patterns.
• ll-issues epic-progress subcommand — Aggregates per-type completion statistics across an EPIC's child issues.
• ll-deps tree --epic subcommand — Renders the EPIC child hierarchy as an ASCII dependency tree. (ENH-1858)
• check_contract FSM evaluator — Detects boundary-mismatch violations where an action crosses a defined scope contract. (FEAT-1791)
• --cascade flag for ll-issues set-status — Propagates EPIC closure to all incomplete child issues.
• Per-state token/cost telemetry in ll-loop run — Records token usage and estimated cost per state transition.
• Session-end sweep hook for stale cross-issue refs — Post-session hook sweeps relates_to: / parent: for references to closed or cancelled issues.
• ll-harness CLI — One-shot runner evaluation CLI for skill/cmd/mcp/prompt with exit-code and semantic criteria. (FEAT-1689)
• ll-history-context CLI — Renders a ## Historical Context block for an issue from .ll/history.db. (ENH-1846)
• Historical context in refine-issue, ready-issue, confidence-check — Prior user corrections and FTS5 matches injected as context. (ENH-1847)
• cli_events table + cli_event_context() — Records every ll- CLI invocation in history.db. (ENH-1848, ENH-1849)
• refine-issue --gap-analysis / --full-rewrite flags — Additive-only gap analysis mode and explicit full-rewrite opt-in. (ENH-1782)
• auto_commit config layer for issues — PostToolUse hook auto-commits issue file writes when enabled. (ENH-1843, ENH-1844)
• analytics.capture config sub-object — Per-category gates for skill, CLI, correction, and file-event capture. (ENH-1840, ENH-1841)
• skill_events table — Tracks /ll: skill invocations as discrete DB events. (ENH-1833)

Fixed

• Log path resolution via cwd field — ll-logs correctly resolves project-relative paths from JSONL records.
• FSM stall detector false positives — exclude_paths prevents bookkeeping writes from masking real stalls. (BUG-1767)
• Session log stores filename not absolute path — Enables cross-machine portability.

Changed

• EPIC critical-path awareness in review-sprint — Surfaces EPIC-blocking issues and warns on absent critical-path items. (ENH-1859)
• Built-in design token profiles in /ll:configure — Scaffolds the selected token profile on first selection.

---

Full Changelog: v1.113.0...v1.114.0