feat: ui

.agent/controller-loop/task.toon

@@ -0,0 +1,9 @@
+engine: "data-driven-v1"
+goal_file: ".agent/controllers/keystone-seam-audit/goal.md"
+plan_file: ".agent/controllers/keystone-seam-audit/plan.toon"
+state_file: ".agent/controllers/keystone-seam-audit/state.toon"
+standards_file: ".agent/controllers/keystone-seam-audit/standards.md"
+branch: "codex/keystone-seam-audit"
+continue_until: "fixed-point"
+max_runs: 12
+max_wall_clock: 4h

.agent/controllers/AGENTS.md

@@ -0,0 +1,18 @@
+# AGENTS.md
+
+## Scope
+
+These instructions apply under `.agent/controllers/`.
+
+## What Belongs Here
+
+- checked-in controller state files
+- checked-in controller plans
+- controller goals and standards documents
+
+## Rules
+
+- Keep controller state in TOON and checked in.
+- Keep plans in TOON and goals/standards in Markdown.
+- Do not add TypeScript, JavaScript, or prompt-template assets here.
+- The Rust engine owns runtime logic; `.agent/controllers/` holds data, not controller code.

.agent/controllers/code-refactoring-act-as-as-a-senior-software-arc/goal.md

@@ -0,0 +1,3 @@
+# Goal
+
+Describe the goal for this controller.

.agent/controllers/code-refactoring-act-as-as-a-senior-software-arc/plan.toon

@@ -0,0 +1,3 @@
+version: 1
+goal_summary: No plan yet
+steps[0]:

.agent/controllers/code-refactoring-act-as-as-a-senior-software-arc/standards.md

@@ -0,0 +1,5 @@
+# Standards
+
+- Keep code maintainable.
+- Avoid one-off hacks.
+- Leave tests green.

.agent/controllers/code-refactoring-act-as-as-a-senior-software-arc/state.toon

@@ -0,0 +1,21 @@
+version: 1
+phase: planning
+goal_status: unknown
+goal_revision: 0
+current_step_id: null
+iteration: 0
+replan_required: false
+completed_steps[0]:
+blocked_steps[0]:
+last_verification: null
+last_cleanup_summary: null
+last_full_test_summary: null
+history[0]:
+notes[0]:
+planning_session:
+  pending_question: null
+  transcript[0]:
+started_at: "1775272586"
+last_usage_refresh_at: "1775272706"
+last_usage_input_tokens: null
+last_usage_output_tokens: null

.agent/controllers/controller-loop/goal.md

@@ -0,0 +1,4 @@
+# Goal
+
+Rewrite `codex-controller-loop` as a Rust TUI-first autonomous controller with TOON-backed machine state and a hard planning/execution phase boundary.
+

.agent/controllers/controller-loop/plan.toon

@@ -0,0 +1,3 @@
+version: 1
+goal_summary: Rust TUI-first autonomous controller
+steps[0]:

.agent/controllers/controller-loop/standards.md

@@ -0,0 +1,8 @@
+# Standards
+
+- Keep the Rust code modular and readable.
+- Treat planning as the only user-input phase.
+- Treat execution as autonomous except for pause, resume, stop, and goal update.
+- Keep controller-owned machine state in TOON files.
+- Leave the codebase in a maintainable state after each completed step.
+

.agent/controllers/controller-loop/state.toon

@@ -0,0 +1,17 @@
+version: 1
+phase: planning
+goal_status: unknown
+goal_revision: 0
+current_step_id: null
+iteration: 0
+replan_required: false
+completed_steps[0]:
+blocked_steps[0]:
+last_verification: null
+last_cleanup_summary: null
+last_full_test_summary: null
+history[0]:
+notes[0]:
+planning_session:
+  pending_question: null
+  transcript[0]:

.agent/controllers/keystone-seam-audit/goal.md

@@ -0,0 +1,20 @@
+# Goal
+
+Identify oversized, hand-maintained source files in the repository, prioritize the highest-value refactor targets, and split them into smaller, cohesive modules without changing external behavior.
+
+A file should be considered a refactor candidate when it is materially large or overloaded, using these default signals:
+- More than 300 lines of hand-written code.
+- Multiple unrelated responsibilities in one file.
+- Difficult-to-test logic mixed with I/O, UI, routing, state wiring, or formatting.
+
+Execution requirements:
+- Ignore generated, vendored, build, cache, and lock files unless the repository clearly treats them as hand-maintained source.
+- Refactor incrementally, one target at a time, starting with the largest safe candidate.
+- Preserve public APIs and user-visible behavior unless a compatibility adjustment is required to complete the split safely.
+- Leave the repository in a clean, test-passing state.
+
+Expected outputs:
+- Smaller files with clearer ownership boundaries.
+- Any necessary import/export or module wiring updates.
+- Tests updated or added when needed to preserve behavior.
+- A concise summary of files split, new module boundaries, and verification results.

.agent/controllers/keystone-seam-audit/plan.toon

@@ -0,0 +1,94 @@
+version: 1
+goal_summary: "Audit the repository for oversized hand-maintained source files, prioritize safe high-value refactor targets, split them into smaller cohesive modules, and finish with passing validation and a clean diff."
+steps[6]:
+  - id: s1
+    title: Establish Safe Refactor Scope
+    purpose: "Load controller inputs and repository constraints, confirm the working tree state, and define the exact file-selection rules so execution can proceed without ambiguity."
+    inputs[4]: ".agent/controllers/keystone-seam-audit/goal.md",".agent/controllers/keystone-seam-audit/standards.md",repository working tree,"repo-level instructions such as AGENTS.md if present"
+    outputs[3]: confirmed execution constraints,candidate exclusion rules,list of locally modified files to avoid or handle carefully
+    dependencies[0]:
+    verification[2]:
+      - label: Check working tree state
+        commands[1]: "git status --short"
+      - label: Locate repo instructions
+        commands[1]: "rg --files -g 'AGENTS.md' -g '.agent/**'"
+    cleanup_requirements[1]{label,description}:
+      No accidental overlap with user edits,Do not modify files with unrelated local changes unless the change is required and the existing edits are understood and preserved.
+    status: active
+    attempts: 1
+  - id: s2
+    title: "Inventory Large Hand-Maintained Files"
+    purpose: "Produce a ranked inventory of oversized source files while excluding generated and third-party material."
+    inputs[2]: repository file list,selection thresholds from goal and standards
+    outputs[3]: ranked candidate list with line counts,"excluded-path list",initial top refactor targets
+    dependencies[1]: s1
+    verification[2]:
+      - label: Enumerate tracked files
+        commands[1]: "rg --files"
+      - label: Rank large files by line count
+        commands[1]: "python - <<'PY'\nimport os, subprocess\nexclude = {'node_modules','dist','build','coverage','.git','.next','.svelte-kit','target','vendor'}\nfiles = subprocess.check_output(['rg','--files']).decode().splitlines()\nrows = []\nfor f in files:\n    parts = set(f.split('/'))\n    if parts & exclude:\n        continue\n    if os.path.splitext(f)[1] in {'.png','.jpg','.jpeg','.gif','.svg','.lock','.snap','.min.js','.map'}:\n        continue\n    try:\n        with open(f,'r',encoding='utf-8') as fh:\n            n = sum(1 for _ in fh)\n    except Exception:\n        continue\n    if n > 300:\n        rows.append((n,f))\nfor n,f in sorted(rows, reverse=True)[:50]:\n    print(f'{n}\\t{f}')\nPY"
+    cleanup_requirements[1]{label,description}:
+      Discard false positives,"Remove generated files, migration dumps, fixtures, and machine-authored artifacts from the candidate list before choosing targets."
+    status: todo
+    attempts: 0
+  - id: s3
+    title: Choose Refactor Order And Boundaries
+    purpose: "Inspect the largest candidates, decide which files are safe and high-value to split first, and define intended module boundaries before editing."
+    inputs[3]: ranked candidate list,current file contents,existing module structure
+    outputs[3]: ordered target list,boundary notes for each target,"explicit non-goals for each refactor"
+    dependencies[1]: s2
+    verification[2]:
+      - label: Inspect top candidates
+        commands[2]: "sed -n '1,220p' <top-candidate-file>","sed -n '221,440p' <top-candidate-file>"
+      - label: Map exports and dependents
+        commands[1]: "rg -n \"from ['\\\"]|require\\(\" <candidate-basename-or-symbols>"
+    cleanup_requirements[1]{label,description}:
+      "Avoid cosmetic-only churn","Do not split files purely by line count; only proceed when coherent seams such as utilities, domain logic, adapters, routes, or components are identifiable."
+    status: todo
+    attempts: 0
+  - id: s4
+    title: Refactor First Target Incrementally
+    purpose: "Split the highest-priority candidate into smaller cohesive files while preserving behavior and keeping the change reviewable."
+    inputs[3]: first target file,boundary notes,repo conventions
+    outputs[3]: new smaller modules,updated imports/exports,"target-specific validation result"
+    dependencies[1]: s3
+    verification[2]:
+      - label: Run targeted tests or checks
+        commands[1]: "<smallest-relevant-test-or-lint-command>"
+      - label: Confirm file size reduction
+        commands[1]: "wc -l <affected-files>"
+    cleanup_requirements[1]{label,description}:
+      Remove temporary seams,"Delete transitional helpers, dead exports, and unused imports created during the split before moving on."
+    status: todo
+    attempts: 0
+  - id: s5
+    title: "Repeat For Remaining High-Value Targets"
+    purpose: "Continue the same refactor pattern for additional oversized files until the main high-value targets are addressed or diminishing returns are reached."
+    inputs[2]: remaining ordered targets,lessons from first refactor
+    outputs[3]: additional split modules,updated dependency wiring,"per-target validation notes"
+    dependencies[1]: s4
+    verification[2]:
+      - label: "Run per-target validation after each split"
+        commands[1]: "<smallest-relevant-test-or-lint-command>"
+      - label: Track remaining oversized files
+        commands[1]: "python - <<'PY'\nimport os, subprocess\nexclude = {'node_modules','dist','build','coverage','.git','.next','.svelte-kit','target','vendor'}\nfiles = subprocess.check_output(['rg','--files']).decode().splitlines()\nfor f in files:\n    parts = set(f.split('/'))\n    if parts & exclude:\n        continue\n    try:\n        with open(f,'r',encoding='utf-8') as fh:\n            n = sum(1 for _ in fh)\n    except Exception:\n        continue\n    if n > 300:\n        print(f'{n}\\t{f}')\nPY"
+    cleanup_requirements[1]{label,description}:
+      Stop at sensible boundary,"Do not keep splitting once modules are cohesive and maintainable; leave well-structured files intact even if they remain moderately large."
+    status: todo
+    attempts: 0
+  - id: s6
+    title: Run Full Validation And Final Cleanup
+    purpose: "Verify repository health, remove leftover refactor debris, and produce a concise execution summary for the controller result."
+    inputs[2]: all refactor changes,repository validation commands
+    outputs[3]: final passing validation results,cleaned diff,"summary of targets, seams, and tests"
+    dependencies[1]: s5
+    verification[2]:
+      - label: Run broadest available validation
+        commands[3]: "<repo-test-command>","<repo-lint-command-if-available>","<repo-build-or-typecheck-command-if-available>"
+      - label: Check for leftover issues
+        commands[2]: "git diff --check","git status --short"
+    cleanup_requirements[2]{label,description}:
+      Leave clean refactor artifacts,"Remove unused files, stale exports, dead code, and temporary comments; ensure only intentional source changes remain."
+      Record outcome,"Summarize which files were split, the new module boundaries, any tests added or updated, and any remaining large files intentionally left unchanged."
+    status: todo
+    attempts: 0

.agent/controllers/keystone-seam-audit/standards.md

@@ -0,0 +1,14 @@
+# Standards
+
+- Prefer cohesive modules with a single clear responsibility over line-count-only splitting.
+- Preserve behavior first; do not mix feature work with the refactor.
+- Avoid one-off abstractions. New modules must reflect real ownership boundaries that the codebase can keep using.
+- Keep public interfaces stable unless a minimal coordinated update is required.
+- Exclude generated, vendored, compiled, snapshot, fixture, and lock files from refactor targets unless explicitly treated as source by the repo.
+- Before editing, check for existing local changes in candidate files and avoid overwriting user work.
+- Refactor in small, reviewable increments so each target can be validated independently.
+- Run the smallest relevant test or lint command after each target, then run the broadest available final validation before finishing.
+- If no automated tests exist for a risky seam, add or strengthen focused tests before or during the refactor when practical.
+- Keep names, module placement, and imports maintainable and consistent with existing repo conventions.
+- Remove dead helpers, obsolete re-exports, and temporary scaffolding introduced during the split.
+- Leave tests green.

.agent/controllers/keystone-seam-audit/state.toon

@@ -0,0 +1,25 @@
+version: 1
+phase: executing
+goal_status: "in-progress"
+goal_revision: 1
+current_step_id: null
+iteration: 0
+replan_required: false
+completed_steps[0]:
+blocked_steps[0]:
+last_verification: null
+last_cleanup_summary: null
+last_full_test_summary: null
+history[0]:
+notes[0]:
+planning_session:
+  pending_question: null
+  transcript[4]{role,content}:
+    user,test
+    assistant,"What should the `keystone-seam-audit` controller actually do? Please specify the target system/repo area, what a \"seam audit\" should inspect, and the expected output artifact(s)."
+    user,find large files and refactor to smaller files
+    assistant,Planning completed
+started_at: "1775273562"
+last_usage_refresh_at: "1775274722"
+last_usage_input_tokens: null
+last_usage_output_tokens: null