MAS-WP-0013 finished: mode contract flows runnable from the bundle alone

conformance/FLOWS.md defines setup, steps, captures and expectations, all
observable at the contract boundary. flows.json carries setups (peer ids),
fixed test keys, statute, scenario and explicit assign steps;
scenarios.json holds the callback scenarios (the stub loads them).
CONTRACT.md defines the case.assign response. Runner reads only the
bundle and rejects undefined keys; new flow s-006 (held over cap).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 5787@bnt-lap001
Assistant-Session: 5db26b82-138b-40cc-9856-4d5eb0f752fd

workplans/MAS-WP-0013-harness-independent-flows.md

@@ -4,7 +4,7 @@ type: workplan
 title: "Mode contract flows independent of the Python harness"
 domain: communication
 repo: mail-address-semantics
-status: proposed
+status: finished
 owner: claude
 topic_slug: coulomb-social
 created: "2026-10-19"
@@ -27,7 +27,7 @@ not run the flows from the bundle alone.
 
 ```task
 id: MAS-WP-0013-T01
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "95fe186c-d0af-5755-a78f-065e7da7afd1"
 ```
@@ -41,7 +41,7 @@ Remove any expectation that only the Python harness can observe.
 
 ```task
 id: MAS-WP-0013-T02
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "2e71569f-739a-513a-970f-0a0c867604d6"
 ```
@@ -49,3 +49,33 @@ state_hub_task_id: "2e71569f-739a-513a-970f-0a0c867604d6"
 Update `mas_modes.runner` and the stubs to read everything from the flow.
 Keep `mas test` green. Add a fixture-level check that the bundle's flows
 do not use undefined keys.
+
+## Outcome
+
+- **`conformance/FLOWS.md`** defines the format: how to set up a run (email
+  side, callback stub `callback-1`, master stub `cms-1`, controllable
+  clock, optional send cap), the five step types, captures and the 14
+  expectation keys. All of it is observable at the contract boundary:
+  messages the stubs receive, command responses, and sent mail.
+- **`flows.json`** carries everything a run needs:
+  - `setups` with role configs by peer id (no URLs; `<mode:ID>` for a
+    primed webhook);
+  - fixed public test keys in `peers`, plus `statute`;
+  - per flow, a `scenario` and an optional `max_sent`;
+  - explicit `assign` steps with `EXT-100` instead of the stub's
+    auto-numbering.
+- **Removed harness-only observables:** the ticket service's response
+  strings (now `event_accepted`, the 2xx) and `tick()`'s return value (now
+  `mode_received`, `mode_received_last` and `mails_sent`).
+- **`conformance/scenarios.json`:** the callback scenarios as data. The
+  Python stub loads them from there.
+- **CONTRACT.md §5** now defines the `case.assign` response, which flows
+  capture from.
+- **New flow** `s-006` (a held send over the cap): 21 flows in all.
+- **Runner** (`mas_modes.runner`) reads only the bundle and rejects
+  undefined keys. `tests/test_modes_conformance.py` checks that every flow
+  uses defined keys, existing setups and scenarios, and that every key is
+  documented in FLOWS.md. The drift test now also covers the vendored
+  FLOWS.md.
+- **Checks.** A mutation check over 5 expectations, plus an undefined key,
+  was detected every time. `mas test`: mode corpus 195/195.