fix: edit-loop guard + self-authored artifact tagging#38

Open

manolitnora wants to merge 171 commits intoHarnessLab:mainfrom

manolitnora:fix/edit-loop-and-self-authored

Commits on Apr 13, 2026

feat(latti): Python lattice mind — solver, streaming, voice

manolitnora
and
claude
committed

Commits on Apr 14, 2026

feat(latti): terminal UI for Claude Code-style formatting

manolitnora
and
claude
committed
chore: add .claw/ to gitignore (session data)

manolitnora
and
claude
committed
feat: allow file operations in additional working directories

manolitnora
and
claude
committed
feat(latti): self-sculpting loop — agent evaluates own output in real-time

manolitnora
and
claude
committed
feat(latti): live self-modification — agent mutates own system prompt in real-time

manolitnora
and
claude
committed
feat(latti): self-optimization — the solver optimizes the solver

manolitnora
and
claude
committed
feat: Lattice class — lattices inside lattices with meet/join/feedback

manolitnora
and
claude
committed
fix: self_optimize.py filler regex + extract_preferences.py eval/selfsculpt extraction
manolitnora
committed
feat(solver): adaptive algorithm selection — scout, classify, pick strategy

manolitnora
and
claude
committed
feat(self-optimize): two-pass scoring — regex + semantic judge via Haiku

manolitnora
and
Latti Nora
committed
fix(voice): three guards prevent voice/chat context mismatch

manolitnora
and
Latti Nora
committed
feat(voice): _detect_llm_spoke wires up the LLM-speaks-first coordination

manolitnora
and
Latti Nora
committed
fix+opt: _detect_llm_spoke reads transcript tool_calls, compile speak regexes once
manolitnora
committed
feat: three turn-completion cues — thinking indicator, done marker, terminal bell
manolitnora
committed
feat: add model_router.py — per-turn model routing with heuristic classifier
manolitnora
committed
feat: wire model_router import into agent_runtime
manolitnora
committed

Commits on Apr 15, 2026

Commits on Apr 16, 2026

Expand lattice solver: sectors, maxent, neural network

manolitnora
and
Latti Nora
committed
Fix TUI colors: brighter green, remove excessive DIM

manolitnora
and
Latti Nora
committed
Add test files and message for Claude Code

manolitnora
and
Latti Nora
committed
Boot hook: gather system state before first LLM call

manolitnora
and
Latti Nora
committed
Latti builder: boot hook, auto-prompt, self-gate, TUI visible boot

manolitnora
and
claude
committed

Commits on Apr 17, 2026

Commits on Apr 19, 2026

latti: add voice completeness guard — reject fragments and incomplete sentences

manolitnora
and
Latti Nora
committed

Commits on Apr 20, 2026

Commits on Apr 22, 2026

Commits on Apr 26, 2026

Commits on Apr 27, 2026

Commits on Apr 28, 2026

Commits on Apr 29, 2026

Citation discipline wired: automatic citation injection on all responses

manolitnora
and
Latti Nora
committed
feat: state machine foundation — typed objects for agent loop

manolitnora
and
Latti Nora
committed
test: state machine integration tests — flag-gated dispatch, surfaces, operators

manolitnora
and
Latti Nora
committed
Fix test import path for agent_state_machine tests

manolitnora
and
Latti Nora
committed
Update state machine comments and tests for Step 6 (2026-04-29)

manolitnora
and
Latti Nora
committed
Implement rotation trigger: pick pending self-axis task when gate fires

manolitnora
and
Latti Nora
committed
Wire typed state machine into agent runtime: bind-on-session, runtime controller, JSON-safe log
manolitnora
committed

Commits on May 1, 2026

Commits on May 2, 2026

fix(state-machine): bind real budget cap to fresh State, not 0.0
manolitnora
committed
feat(identity): v1c — mark natural-language fake refs (Decision #N etc)
manolitnora
committed
Allow forced supervisor smoke

manolitnora
and
Latti Nora
committed
Show Latti control-plane modes in status

manolitnora
and
Latti Nora
committed

Commits on May 3, 2026

Commits on May 4, 2026

Commits on May 5, 2026