Serialize TaskStore writes

…e validation

RED phase for Plan 02-08 cycle-detector + envelope-enforcement invariants:

tests/test_cycle_detector.py (7 tests):
- no-cycle on single message (negative case)
- 3 A->B + 3 B->A round-trips in last-20 window trip cycle (D-20/D-21)
- window bounded to 20 — older round-trips don't count
- topic-hash priority chain: dedupe_key -> request_id -> sha1(content[:128]) (D-19)
- suppressedTopics persist on disk; re-entry deduped as cycle_suppressed_existing
- Pitfall HKUDS#3: progressSignal entry breaks streak (skipif Plan 02-09 not merged)
- Pitfall HKUDS#7: sync emit — handler observes suppressedTopics on disk when fired

tests/test_transport_envelope.py (6 tests):
- _pre_deliver_hooks parses JSON bytes -> TeamMessage
- BC pass-through: no envelope fields -> no validation (Pitfall HKUDS#8)
- Partial envelope -> MalformedEnvelopeError
- Full valid envelope -> accepted
- 8 consecutive malformed -> DriftRegression (D-09)
- Valid envelope resets per-agent counter

RED confirmed: cycle_detector 4 failed + 2 passed + 1 skipped;
transport_envelope ImportError on _pre_deliver_hooks / _reset_drift_counters.

…n Transport.deliver

GREEN phase — implements the two Phase 2 hook sites that break A<->B deadlock
loops and enforce the envelope protocol scope-limited for BC.

DefaultRoutingPolicy (clawteam/team/routing_policy.py):
  - decide() prefixed with cycle detection BEFORE existing throttle (D-18);
    reuses the existing recentEvents 50-entry bounded window (specifics
    lesson HKUDS#1, no parallel tracker).
  - _topic_hash(envelope): priority chain dedupe_key -> payload.request_id ->
    sha1(content[:128])[:16] (D-19).
  - _detect_cycle(state, envelope, topic_hash): scans last 20 recentEvents for
    >=3 A->B AND >=3 B->A with matching topic_hash (D-20); returns None when any
    entry carries progressSignal: True (Pitfall HKUDS#3 mitigation).
  - Cycle hit persists topic_hash onto routes[route_key].suppressedTopics, saves
    state BEFORE sync emit (Pitfall HKUDS#7), then emits CycleDetected with pair,
    topic_hash, route_keys, window_size=20.
  - Subsequent hits for same route+topic short-circuit with
    reason='cycle_suppressed_existing' (no re-emit, D-21).
  - _append_event gains topic_hash kwarg; pending/aggregated/cycle_suppressed
    entries now carry topicHash so future decide() calls can correlate.

Transport (clawteam/transport/base.py):
  - _pre_deliver_hooks(data) -> TeamMessage: parses JSON bytes, short-circuits
    when msg has none-of-three envelope fields (Pitfall HKUDS#8 BC pass-through),
    otherwise requires all three via TurnEnvelope.model_validate.
  - Per-agent module-level consecutive-malformed counter (D-09); threshold = 8
    consecutive strikes emits DriftRegression (sync). Any valid envelope resets
    the counter for that agent. MalformedEnvelope emitted on every strike.
  - _reset_drift_counters(): test + SprintConductor-teardown helper (Plan 02-11
    contract).
  - Transport ABC unchanged (public interface preserved).

FileTransport.deliver (clawteam/transport/file.py):
  - Invokes _pre_deliver_hooks at the top; MalformedEnvelopeError re-raised to
    caller so MailboxManager can route to dead-letter / surface to the agent.
  - p2p.py falls back to FileTransport.deliver so its delivery path inherits
    the hook automatically; no direct change needed there for v1.

Tests: 698 passed, 1 skipped (Plan 02-09 gate); ruff clean.
Regression matrix (12 templates) and test_runtime_routing.py both still green.

Adds plan-level SUMMARY covering Task 1 (RED) + Task 2 (GREEN) for the
cycle detector in DefaultRoutingPolicy.decide() and the envelope-validation
pre-deliver hook in Transport.deliver.

Key outcomes recorded:
- D-18..D-21 cycle detection (tail-20 window, 3+ round-trip threshold)
- D-19 topic-hash priority chain (dedupe_key -> request_id -> sha1)
- Pitfall HKUDS#3 progressSignal short-circuit (populator lands in Plan 02-09)
- Pitfall HKUDS#7 sync emit after _save_state
- D-08/D-09 envelope enforcement at Transport.deliver() with Pitfall HKUDS#8 BC
- 8-strike DriftRegression + per-agent counter reset on valid envelope

Forward contracts for Plan 02-09 (progressSignal populator) and Plan 02-11
(SprintConductor _reset_drift_counters teardown) documented.

…n tests

- EXISTING_TEMPLATES_PHASE3 canonical list + len==6 guard (T-09-02)
- gstack_plugin_loaded fixture: fresh GstackSprintPlugin per test, minimal
  fake ctx, no EvidenceSchemaRegistry touch (avoids test-ordering collision
  with tests/test_gstack_plugin.py)
- test_existing_template_spawn_unaffected_by_gstack_plugin[6 templates] —
  asserts TeamConfig.template != 'gstack', member roster != 11 gstack roles,
  leader_role != 'ceo' for each of software-dev/hedge-fund/code-review/
  harness-default/research-paper/strategy-room (Pitfall 14 + QUALITY-14 +
  success-criterion HKUDS#2)
- test_reflect_phase_does_not_trigger_phase6_pending_write_for_non_gstack
  [6 templates] — T-07-01 HIGH severity 2nd-layer mitigation anchor;
  runtime assertion that emitting a Reflect event for non-gstack teams
  never writes _phase6_pending/<sprint>-retro.json
- test_all_six_existing_templates_parse_with_gstack_loaded — aggregate
  assertion that each existing template's phases != gstack phases AND
  leader_role != 'ceo' (success-criterion HKUDS#7)

Full-suite delta: +13 new passing tests (958 -> 971 passed). 3 pre-existing
failures unchanged (documented in deferred-items.md).

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>