hermes/automaton
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
main
automaton/tasks/state-file-enforcement/IMPLEMENTATION.md
T
Lap Tran bc7daf8590 Restore archived tasks, fix dashboard scroll-reset, bind ornith, add Playwright smoke test 
- **Restore 82 completed tasks** from tasks/complete/ back to tasks/ top
  level (all <7 days old per the cleanup policy; premature bulk archive
  was fixed).
- **Dashboard: fix scroll-reset on auto-refresh** — renderBoard rebuilds
  the board via innerHTML every 2s, destroying each column-body's
  scrollTop. Now snapshots column-body scrollTop + board.scrollLeft +
  view.scrollTop before rebuild and restores after (matched by
  PHASE_GROUPS index).
- **Dashboard UI additions** (pre-existing unstaged work): approval
  section cards, transition buttons, inline artifact editor (textarea for
  writing missing SPEC/VERDICT/etc from the detail modal).
- **Bind ornith as Implement model** — config.md: Model explicit to
  omlx/Ornith-1.0-35B-4bit-mlx, context window 32768. Interactive
  autopilot already used ornith via opencode default; now explicit.
- **Fix cleanup stub** — automaton-cleanup.sh had a stale --project arg
  pointing at a pytest temp dir (test isolation leak). Rewired to point
  at ~/.automaton.
- **Fix plist-isolation test** — test asserted host plist doesn't exist,
  but a real install creates it. Now snapshots mtime before run, asserts
  unchanged after (only a write during the test counts as bleed).
- **New Playwright smoke test** (tests/test_dashboard_ui.py) — 2 tests:
  board renders tasks, column scroll survives auto-refresh tick.
  Verified the test fails without the scroll fix (scrollTop resets to 0).
  Skipped via importorskip when playwright is absent (main CI stays
  green).
- **Clarify SI loop scope in README** — new-project onboarding section
  documents the framework-scoped self-improvement loop and options
  (leave/pause/create project loop).
- **CHANGELOG** documents all changes including the known model-divergence
  gap (mde tasks marked complete but per-role model binding was never
  implemented).
2026-06-26 10:05:18 -04:00

2.6 KiB
Raw Permalink Blame History

Implementation: State File Enforcement

Changes Made

1. .state file format (implemented in scripts/status.py)

  • .state file contains a single phase name (e.g., research, research:awaiting_approval, implement)
  • Written atomically via .state.tmp → .state rename
  • .state.approvals append-only log records all approvals with timestamp and approver
  • .state.lock for multi-agent claiming (optional, only in multi-agent mode)
  • Non-artifact metadata files (.state, .state.approvals, .state.lock, .state.tmp, VRAM_CONFIG.md, PARENT_SPEC.md, REVIEW.md) are excluded from artifact checks

2. State transitions with approval sub-states

  • Approval-gated phases: research, decomposition, design, test_design now have :awaiting_approval → :approved sub-states
  • Non-approval phases: implement, bug_find, adversarial_bug_find, doc_review, referee have no sub-states
  • status.py --transition refuses transitions past :awaiting_approval without --approve
  • status.py --approve transitions :awaiting_approval → :approved and records approval in .state.approvals

3. Backward compatibility

  • If .state doesn't exist, status.py infers phase from artifacts and writes .state
  • upgrade.sh bootstraps .state for all existing tasks
  • Phase prompts work with or without .state (warns if missing)

4. Forbidden artifacts per phase (implemented in status.py --validate-folder)

  • Each phase has a defined set of artifacts that must NOT exist (artifacts from future phases)
  • --validate-folder checks and reports violations
  • --transition refuses to proceed if forbidden artifacts exist

5. Task creation gate (implemented in status.py --create-task)

  • --create-task creates task folder with .state = new and empty .state.approvals
  • Validates kebab-case task names
  • Refuses if task already exists
  • --audit Category 4 flags manually created task folders

6. Orchestrator and workflow updates

  • prompts/workflow.md rewritten: .state is canonical, approval sub-states documented, status.py commands referenced
  • prompts/orchestrate.md reduced from 493 to 143 lines, references workflow.md and subtask_management.md
  • All phase prompts include .state precondition check

Files Modified

  • scripts/status.py (new, 980 lines)
  • prompts/workflow.md (rewritten)
  • prompts/orchestrate.md (rewritten, 143 lines)
  • prompts/subtask_management.md (new, extracted from orchestrate.md)
  • tests/test_status.py (new, 25 tests)
  • scripts/upgrade.sh (new)

Test Results

  • 183 tests passing (including 25 new status.py tests)
  • Python compilation clean
  • Shell script syntax clean
Reference in New Issue View Git Blame Copy Permalink