hermes/automaton
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
main
automaton/tasks/task-status-reason/ADVERSARIAL_BUG_REPORT.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

2.9 KiB
Raw Permalink Blame History

Adversarial Bug Report: task-status-reason

Summary

Deep-dive audit found 3 subtle issues: misleading red border on bug_find cards, redundant verdict parsing, and no status_reason propagation to subtasks.

Bugs Found

Bug 1: Red error border on bug_find task cards is misleading

  • Severity: Low
  • Location: automaton/dashboard/html/styles.css — .task-card-reason rule
  • Description: The .task-card-reason CSS uses border-left: 2px solid var(--error) for all three states (blocked, bug_find, adv_bug_find). But bug_find and adv_bug_find are normal workflow states, not errors. A green or neutral border would be more appropriate for non-terminal states.
  • Reproduction: Open any task in bug_find or adv_bug_find state — the reason banner shows a red border suggesting something is wrong, when it's expected behavior.
  • Suggested Fix: Use var(--warning) for bug_find states and reserve var(--error) for blocked only.

Bug 2: Redundant verdict parsing on every status_reason access

  • Severity: Low
  • Location: automaton/dashboard/core/task.py:136-175 — status_reason property
  • Description: status_reason calls parse_verdict_status() again for BLOCKED/DONE tasks, even though determine_task_state() already parsed the verdict to classify the state. For DONE, it doesn't re-parse (just returns "Verdict: PASS"), but for BLOCKED it re-reads the verdict content and re-parses. This is redundant but cheap given the small number of tasks.
  • Reproduction: Every time the dashboard renders, BLOCKED tasks trigger a second verdict parse just for the display string.
  • Suggested Fix: Cache the parsed verdict status on the Task object during discover_tasks(), e.g., a _verdict_status field that status_reason can reference instead of re-parsing.

Bug 3: Subtask state not visible in detail panel reason

  • Severity: Low
  • Location: automaton/dashboard/html/dashboard.js:236-241 — subtask list in detail panel
  • Description: Subtasks in the detail panel show only a verdict pass/fail indicator. Unlike parent tasks, blocked subtasks don't display their status_reason in the list. A blocked subtask just shows "✗" next to its name with no explanation of why.
  • Suggested Fix: For blocked subtasks, include the subtask's verdict status (FAIL/NEEDS_REVIEW) in the list item, or hover tooltip with the reason.

Bug 4: Fallthrough status_reason never reached, masks missing states

  • Severity: Low
  • Location: automaton/dashboard/core/task.py:175
  • Description: The fallthrough return f"In {self.state.value} phase" is dead code — every TaskState value is covered by explicit branches. If a new state is added (e.g., INTEGRATION_TEST), no compile-time error occurs and a vague message is shown. Contrast with Python enums which have no exhaustiveness checking.
  • Suggested Fix: Add # pragma: no cover or raise/log a warning if a new state goes unhandled.

Score

+3

Reference in New Issue View Git Blame Copy Permalink