fix: improve error handling for config loading, task parsing and file transport

…st_command + deploy_url

- 13 failing tests cover presence, frontmatter schema, stub detection (sections + blacklist), test-report rerun + cache-hit + exit-1 + timeout, and ship-notes deploy_url 2xx/non-2xx/URLError paths
- Test-local schema fixtures (DesignDoc/TestReport/ShipNotes) inherit from ArtifactFrontmatterBase; Phase 3 GstackSprintPlugin ships the real subclasses
- Pitfall HKUDS#4 proof: subprocess_runner injection seam asserted with shell-free list cmd + cwd pin + 300s timeout
- Pitfall HKUDS#6 proof: unregistered artifact_type surfaces 'Phase 3 plugin must register' message
- Pitfall HKUDS#10 proof: deploy_url_checker injection seam covers URLError offline-safety path
- RED: all tests fail with ModuleNotFoundError on clawteam.harness.evidence_gate

…oy_url HEAD probe

- New clawteam/harness/evidence_gate.py subclasses ArtifactRequiredGate with:
  - Layer 1 presence check (inherited)
  - Layer 2 frontmatter parse + EvidenceSchemaRegistry dispatch; unregistered artifact_type surfaces 'Phase 3 plugin must register'
  - Layer 3 stub detection (detect_stub): blacklist regex (TBD / TODO / xxx+ / placeholder / Lorem ipsum) with line numbers, then required ## section floor of 100 bytes
  - Layer 4 dispatch: test-report → _run_test_command (shell=False, cwd=workspace_branch, 300s timeout) with SHA256 content-hash cache persisted via file_locked + atomic_write_text; ship-notes → _check_deploy_url HEAD probe with 10s timeout
  - Layer 4b per-phase artifact-cap (phase_artifact_cap_bytes) with artifact_compaction suggestion
- Injection seams: subprocess_runner (T-02-03 mitigation verification) and deploy_url_checker (Pitfall HKUDS#10 / T-02-01 Phase 5 allowlist contract)
- Fix test file: rename TestReportFixture to ReportTestFixture (avoid pytest Test* collection warning); quote created_at YAML values so they stay str (YAML 1.1 auto-parses bare ISO-8601 as datetime)
- 13 evidence-gate tests pass; template regression matrix (SC#10), harness, artifact_caps, evidence_schemas, turn_envelope all green (87/87)

- ToolCallCompleted (D-09, cost tracker input)
- ClaudeApiResponse (D-12, cache-hit tracking)
- BudgetAlarmReached (D-10, budget threshold crossing)
- RateLimitSaturated (D-13, 429 backpressure)
- ZombieWorktreeGced (SC HKUDS#10, worktree GC)
- DormancyTransition (QUALITY-04, agent dormancy)
- register_event_type calls appended at module bottom after Phase 6 block

…2 GREEN)

- clawteam/cli/commands.py doctor() gains optional `gc: bool = typer.Option(...)`
  flag. When `--gc` absent, behavior is byte-identical to prior doctor()
  (existing _DOCTOR_TOOLS iteration + _human() render + _output(checks, _human)).

- With `--gc`: after the tool checklist, walks get_data_dir()/teams/<name>/
  for each team dir. For each, collects active_branches from every
  SprintState.load(team, sprint_id) under sprints/ (unreadable state.json
  fallbacks to sprint_dir.name — preserve-on-doubt safety posture).
  Calls find_zombie_worktrees(worktrees_root, max_age_days=30,
  active_branches=active_branches) then gc_zombies(zombies, team_name=...,
  bus=get_event_bus()) emitting one ZombieWorktreeGced per deletion.
  Appends per-team zombies_removed + active_branches_preserved + disk_usage
  report to gc_summary["teams"]; combined dict {**checks, "gc": gc_summary}
  dispatched via _output for both --json and human branches.

- Human renderer extension: after the existing tool-checklist render,
  prints "GC complete: N zombie worktree(s) removed across M team(s)"
  then per-team line with disk_usage gb + over_soft/over_hard color tag
  (green ok / yellow over-soft / red over-hard).

- Safety invariant (07-RESEARCH §Pitfall 6): active SprintState
  workspace_branches preserved regardless of worktree age. Test
  test_doctor_gc_respects_active_sprints locks this contract.

Tests: tests/test_doctor_gc.py 4 passed + tests/test_doctor.py 6 passed
+ tests/test_doctor_new_entries.py 22 passed + tests/workspace/ 13 passed
= 45 passed total (BC + new, no regressions).

Closes Plan 07-08 Task 2; ROADMAP §Phase 7 SC HKUDS#10 (zombie worktree GC)
closed for the doctor-entry half. Wave 5 Plan 07-08 complete.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

- .planning/phases/07-parallel-sprints-attentionqueue-ux-cost-controls/07-08-SUMMARY.md (new)
  Documents both tasks, 2 deviations (Rule 3 cross-executor stash on
  Task 1 GREEN + Rule 1 fixture mtime bump), interface surface for
  Plan 07-09, 17 tests green, self-check passed.
- .planning/STATE.md status/stopped_at/Current Position/session/decisions/metrics
  advanced for 07-08; completed_plans 73 -> 74 (96%).
- .planning/ROADMAP.md Phase 7 plans table marks 07-08 complete; progress
  row flipped from 6/9 (Wave 4) to 8/9 (Wave 5 — only 07-09 load test remains).

ROADMAP §Phase 7 SC HKUDS#10 (zombie-worktree auto-GC + 5 GB/10 GB disk budget)
CLOSED for the doctor-entry half. Sprint-spawn-time auto-GC deferred to
v1.x per scope boundary.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The Phase 7 GC pass previously scanned teams/<team>/worktrees/ — a
directory WorkspaceManager never writes to. Real worktrees live at
<data_dir>/workspaces/<team>/<agent>/ (workspace/manager._workspaces_root).
Net effect: every production invocation of clawteam doctor --gc
reported total_zombies=0 even with real zombies on disk; Phase 7 SC HKUDS#10
was silently unmet. Exposed workspaces_root() as a public accessor,
switched the CLI scan root to workspaces/<team>/, and rewrote
tests/test_doctor_gc.py to build worktrees under the production layout
so the test now exercises the shipped path rather than a parallel one.
Team discovery now unions teams/ and workspaces/ directories so stale
workspaces without a matching teams/ entry are still GC'd.

Active-branch safety filter refinement to active_agents (IN-07) is
deferred; this commit fixes the scan path only.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

User request: "用tmux hook 把prompt注入进每一轮对话 包括身份和需要用的
skill 或者用tmux hook去做一些别的事情 总之tmux hook会是一个很好的技术".

Concept: agents drift over long sessions. CEO forgets to delegate (user
just hit this bug), engineer forgets which `clawteam` subcommands are
available, shipper forgets that `/ship` lives in the framework. A
periodic re-injection of "you are X; here's what you should use"
keeps them anchored.

Implementation:

1. `clawteam nudge <pane_id>` CLI command (new clawteam/nudge.py):
   - Looks up pane_id → session → team → agent via the persisted
     tmux_pane_map.json
   - Debounces (60s cooldown per pane) so repeated silence-hook firings
     don't spam the claude TUI
   - Reads the latest sprint state (current_phase + goal) for context
   - Builds a role-specific reminder text from _ROLE_NUDGES table
     (each role gets 2-3 lines: what you ARE + what commands you CAN
     run now). Appends "Goal: X · Phase: Y" when available.
   - Injects via `tmux send-keys -l <text>` + `send-keys Enter` so
     claude processes it as a user turn.
   - Silent-fail on any error (a noisy hook drowns the user's tmux).

2. tmux hook wiring (tmux_backend.py::_configure_nudge_hook):
   - `monitor-silence = 45s` on each window
   - `alert-silence` hook fires `clawteam nudge #{pane_id}` in the
     background (`run-shell -b`)
   - Opt-out via `CLAWTEAM_NUDGE_SECONDS=0` env var
   - Called from both `TmuxBackend.tile_panes()` (happy path) and
     `TmuxBackend.enable_mouse()` (the --windows non-tiled path) so
     every team gets nudge wiring regardless of layout choice.

3. pane_map normalization (tmux_backend.py::tile_panes):
   - Post-layout readback of pane_title returned claude's dynamic
     status strings like "⠐ CEO" / "✢ ENG-MGR" (spinner glyphs prefix).
   - Normalize by matching uppercase title content against the
     pre-merge window roster (roster_upper dict). Any title containing
     a known role (e.g. "⠐ CEO" contains "CEO") maps to the canonical
     lowercase role name.
   - pane_map.json now reliably stores {"0": "ceo", "1": "pm", ...}
     regardless of what claude's currently showing.

Real-env verified:
  clawteam go --no-attach --no-window "test"
  cat ~/.clawteam/teams/<team>/tmux_pane_map.json  → clean lowercase roster
  clawteam status                                    → Panes: 0=ceo 1=pm ...
  # wait for claude to boot
  clawteam nudge --force <pane_id>                   → injects role reminder
  tmux capture-pane                                  → sees "[nudge] Reminder:
                                                      you DELEGATE, never
                                                      implement. Decompose..."

Future uses of this hook infrastructure (brainstormed but not impl):
  - HKUDS#3 pane-died → log death reason, optional respawn
  - HKUDS#5 alert-activity → append pane output to activity.jsonl timeline
  - HKUDS#6 session-created → broadcast phase+goal to all panes
  - HKUDS#7 send-keys filter → intercept dangerous commands (rm -rf, force push)
  - HKUDS#10 pipe-pane → validate agent envelope JSON against schema

Tests: 30 solo/spawn_cli still green (nudge unit tests not added yet
— next commit candidate).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>