Roles, capability profiles and help on the new loop (M2.6)

- roles.driver (unset: the planner's model) and roles.classifier (unset:
  the expert's); the mapping is pinned when an item opens, and the driver
  and the go-ahead reader run on the item's own (L9).
- profiles: in configuration (tools, strengths, weaknesses), used in
  briefings: the driver's instructions say what its helper and expert are
  good at, the helper's brief who reads its handback; never a model name.
- delegate(task) and ask_expert(question) for the driver. A helper works
  on a branch off the milestone's; on a handback its work comes back as
  uncommitted changes, otherwise it is discarded, so harness branches
  never hold commits only the clone has. ask_expert keeps its budget.
- Triggers: failure_limit failed commands in a row (Driver's new "stuck"
  outcome) or the turn cap, mid-milestone, call a helper in once per
  milestone; if it fails too, the item waits on the operator. A
  delegation cut off by a crash is discarded at boot.
- Evaluation capture: delegations and ask_expert as before, and the
  reader's verdicts as classifier pairs.

Tests: help tests (merge-back, discard, needs a milestone, the stuck
trigger and its once-per-milestone limit, the expert and its budget,
pinning, briefings by profile, boot discarding a cut-off delegation),
config tests for roles and profiles, and the walk delegating with a
scripted helper that sometimes fails, with a delegations_settle property.
The walk found helper commits stranded on a merged PR's branch, fixed by
returning a helper's work uncommitted.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE

nisse-dev/tests/test_loop_properties.py

@@ -23,10 +23,13 @@ from test_loop_projects import Router
 
 from nisse_dev.cancel import Cancel
 from nisse_dev.canonical import Fact, PlanSubmitted, ToolCall, UserMessage
+from nisse_dev.delegation import DELEGATION_INTENT, Caps
 from nisse_dev.execution import SSHHost
 from nisse_dev.gitea import Comment, Issue
+from nisse_dev.llm import RoleMap
 from nisse_dev.loop.agent import LoopAgent, Project
 from nisse_dev.loop.driver import Driver
+from nisse_dev.loop.help import Helpers
 from nisse_dev.loop.items import (
     ANSWER,
     BLOCKED,
@@ -97,6 +100,8 @@ class PolicyModel:
             return [call("ask_human", question=f"talk first {self.n}?")]  # discuss
         if "finish_milestone" not in offered:
             return [call("restate", text=f"restatement {self.n}")]
+        if "delegate" in offered and self.rng.random() < 0.1:
+            return [call("delegate", task=f"help {self.n}")]
         if "wait_for_review" in offered and self.rng.random() < 0.2:
             return [call("wait_for_review", note=f"waiting {self.n}")]
         if self.rng.random() < 0.15:
@@ -120,6 +125,32 @@ class PolicyModel:
         ]
 
 
+class HelperModel:
+    """A scripted helper: does some work, then hands back, succeeding or
+    not as the seed says."""
+
+    def __init__(self, rng: random.Random) -> None:
+        self.rng = rng
+        self.n = 0
+
+    def complete(self, events, tools, model, system=None) -> list[Any]:
+        assert_renderable(events)
+        self.n += 1
+        return [
+            call("write_file", path=f"h{self.n}.py", content=f"h = {self.n}\n"),
+            ToolCall(
+                call_id=f"hb{self.n}",
+                tool="hand_back",
+                arguments={
+                    "changed": f"h{self.n}.py",
+                    "now_true": "helped",
+                    "next": "go on",
+                    "succeeded": self.rng.random() < 0.6,
+                },
+            ),
+        ]
+
+
 class Walk:
     """``worlds``: project name -> its world (forge, bare repo, clone,
     milestones), all over one store."""
@@ -174,13 +205,18 @@ class Walk:
             cancel=cancel,
             oplog=OpLog("nisse-dev", _Null()),
             reader=self,  # type: ignore[arg-type]
+            helpers=Helpers(
+                client=HelperModel(self.rng),  # type: ignore[arg-type]
+                roles=RoleMap(executor="", planner="m", helper="h", expert="x"),
+                caps=Caps(turns=4),
+            ),
             bug_grace_s=600,
             clock=lambda: self.now,
         )
         transport.on_message(agent.on_message)
         return agent, transport
 
-    def agrees(self, proposal: str, reply: str) -> bool:
+    def agrees(self, proposal: str, reply: str, model: str = "") -> bool:
         return self.reads_yes(proposal, reply)
 
     def items(self) -> list[Any]:
@@ -404,7 +440,22 @@ class Walk:
                 assert not (loop.prepared or loop.milestone or loop.pr), (
                     f"conversations_do_no_work: {item.id}: {where}"
                 )
+        # L9: a delegation never outlives its step: merged back or discarded.
+        assert not [
+            i for i in store.pending_intents() if i.kind == DELEGATION_INTENT
+        ], f"delegations_settle (intent): {where}"
         for name, w in self.worlds.items():
+            helping = [
+                b
+                for b in [
+                    *w["repo"].local_branches(),
+                    *_git(
+                        w["bare"], "for-each-ref", "--format=%(refname:short)"
+                    ).split(),
+                ]
+                if "-help-" in b
+            ]
+            assert not helping, f"delegations_settle (branch {helping}): {where}"
             self._records_have_one_item(name, w, where)
             self._own_repository_only(name, w, where)
             self._parked_work_safe(w, name, where)
@@ -519,7 +570,10 @@ class Walk:
             written = {
                 str(e.event.arguments.get("path"))
                 for e in self.world["store"].transcript(item.id, "driver").read()
-                if isinstance(e.event, ToolCall) and e.event.tool == "write_file"
+                # The driver's own: a failed helper's are discarded on purpose.
+                if isinstance(e.event, ToolCall)
+                and e.event.tool == "write_file"
+                and e.actor == "driver"
             }
             here = current.startswith(f"{item.id}-")
             for path in written: