v0.30.0

Features

• Monitor shadow zone (t1216_2): The full monitor now shows the shadow agent side by side with the agent it follows, with its own live preview, key forwarding and a grace window so a brief capture blip doesn't make the panel vanish.
• Monitor concern picker (t1216_3): Shadow concerns now surface in the full monitor with a per-agent badge, an automatic toast when new concerns appear, and a c picker for reviewing them.
• Monitor shadow spawn (t1216_4): The e / E shadow-spawn keys work in the full monitor, not just the minimonitor, with a duplicate guard so you can't accidentally launch a second shadow for the same agent.
• Syncer version tab (t1223_3): The syncer gained a Versions tab that shows the framework version installed in each linked repo and can upgrade them, including a clean self-handoff when the syncer upgrades the repo it is running from.
• Cross-repo settings seam (t1223_4): Configuration can now be merged into another repo safely — writes are atomic, type conflicts are reported instead of silently overwriting, and a partial import tells you exactly what landed.
• Syncer settings tab (t1223_5): A new Settings tab shows how each configuration value resolves across your linked repos and adds a four-step wizard for pushing one setting to other repos, with masking for sensitive values.
• Pick a task by number (t1310): Press p in the minimonitor to launch a task by typing its number, with a confirmation dialog that warns you about unmet dependencies first.
• COMPLETED agent status (t1322): The monitor and minimonitor now distinguish agents that finished their task from those that are merely idle, with a dedicated badge, counter and auto-switch filter.
• Cross-repo agent marks (t1326): Press space in either monitor to mark an agent as prioritized; marks are shared across repos, so an agent you flag in one project stays flagged everywhere.
• Release docs-gap skill (t1355): A new aitask-docs-gap skill reviews everything landed since the last release, classifies which changes are missing documentation, and files a single documentation task covering the gaps.

Bug Fixes

• Python suite verdict (t1179): The Python test suite now ends with an unmissable pass/fail banner derived from the real exit status, so a failing run can no longer look like a passing one.
• Codex skill launch (t1221): Launching a skill through Codex now validates its arguments across every operation and refuses malformed input instead of composing a broken command.
• Verified-parity baseline (t1232): Fixed a test that compared shipped model scores against live project config, so local model customizations no longer break the suite.
• Board field persistence (t1243_2): Saving a board change now writes only the fields the caller actually changed, instead of rewriting the whole set and clobbering concurrent edits.
• Board filter row width (t1247): The board's filter row now measures its own width and reflows the search box vertically on narrow terminals instead of overflowing.
• Board scroll jump (t1248): Navigating the board no longer jumps unexpectedly when the focused card is off-screen — movement is re-anchored to what you can actually see, and focus scrolls immediately.
• Plain-bullet verification checklists (t1264): A verification checklist written with plain bullets is no longer unrecoverable — a new convert action turns them into proper checkbox items without losing text or indentation.
• Silent push failures (t1265): ait git push now reports when a task-data push actually pushed nothing, instead of exiting successfully and leaving you to discover it later.
• By-Trail refresh (t1268): The By-Trail board view gained a clear refresh ladder (local, freshness, agent), per-card drift markers and accurate per-view key/footer labels.
• Board header row (t1278): The board's header row is drawn again, and the trail freshness marker survives resizing instead of being truncated away.
• Double-tapped refresh (t1279): A double-tapped R on the By-Trail agent refresh no longer fires the dialog twice.
• Followed-agent task info (t1282): A new I shortcut in the minimonitor opens task info for the agent you are following, even when a different card has focus.
• Board error handling (t1302): The board now degrades gracefully on any filesystem error during refresh, and board keys keep a stable position in task frontmatter instead of shuffling on every save.
• Shadow review gate (t1311): Fixed the shadow's implementation-review gate, which previously refused to review on a false "too early" premise, and added a configurable review depth per profile.
• Board dialog errors (t1314): Board dialogs no longer crash on unexpected subprocess errors, and the loading overlay is always dismissed.
• Skill surface guards (t1325): The skill verification guards no longer validate themselves against their own output, so a missing skill surface is actually caught.

Improvements

• Gate registry sync (t635_34): A new ait gates sync-registry reconciles an installed project's gate registry with the framework reference — filling in missing fields, reporting conflicts and preserving your comments — plus a warning at pick time when an active gate has no verifier configured.
• Shared shadow seam (t1216_1): The shadow capture and snapshot logic moved out of the minimonitor into shared code so both monitors behave identically, with proper ordering guarantees when a shadow is rebound.
• Configurable merge target (t1233): A new output_branch profile key lets a task's final merge target be something other than your default branch, recorded in the plan header and drift-checked before the merge.
• Stats module promotion (t1235): Statistics data loading moved to the shared library layer, removing a path hack that made the work-report tooling reach into the stats TUI.
• Syncer arrow navigation (t1266): Arrow keys now move between the syncer's tab bar and its content, so you can navigate the whole TUI without reaching for other keys.
• Shared narrow-terminal breakpoint (t1251): The narrow/normal/wide terminal breakpoints are now defined once and shared, so TUIs adapt consistently at the same widths.
• Structured concern dispositions (t1274): Shadow concerns are now grouped by disposition in the picker, informational items are dimmed, untitled rows show a placeholder, and unrecovered markers are called out instead of silently dropped.
• Shadow pane-id hardening (t1307): The shadow skill no longer breaks when a pane id is truncated, and it can recover the right target through its own binding or an exact-match lookup.
• Explore label handling (t1312): Labels suggested during explore are now added to your project vocabulary and confirmed with you first, with a new profile key controlling whether that happens automatically, on request, or not at all.

Documentation

• Syncer documentation (t1223_6): The syncer docs now cover all three tabs, the arrow-key navigation model, framework-version inspection and upgrades, and the cross-repo settings push wizard.
• Docs gaps since v0.29.0 (t1361): Added a full reference page for the ait gates command family and documented the By-Trail board view, the COMPLETED agent status and the verification-checklist convert action.

Performance

• Board test fixture harness (t1354_1): A shared board test fixture replaces per-file setup in the slowest board test files, cutting suite runtime and removing tests that used to skip themselves.

Tests

• Isolated Python test lane (t1236): The Python test runner no longer leaks its own import path into tests, and a new lane imports each test file in a clean interpreter to catch hidden bootstrap assumptions.
• Board movement baseline (t1243_1): Added a characterization suite and performance baseline for board card movement, pinning current behaviour before reordering work begins.
• Restored isolation lane (t1306): Recovered the Python isolation lane and supporting files that a history rewrite had dropped.
• Skill dispatch smoke test (t1317): A new test checks every templated skill against all four agent surfaces, and the missing OpenCode stub it found was added.
• Derived agent defaults (t1318): Tests now derive expected code-agent defaults instead of pinning model names, so adding a model no longer breaks unrelated suites.
• Label frontmatter characterization (t1321): Pinned how --batch --labels normalizes, dedupes and sanitizes labels across all three task-creation paths.
• Live shadow spawn smoke (t1353): A live isolated-tmux smoke test now exercises monitor shadow spawn end to end against a real tmux server.