Merge pull request 'Add role-boundary proofs for reviewer queue workflows (Issue #175)' (#192) from feat/issue-175-role-pivot-wall into master

review_proofs.py

@@ -330,9 +330,97 @@ def assess_self_review_contamination(reviewer_identity, pr_author,
     }
 
 
+def assess_role_boundary(proof):
+    """Assess reviewer/author role separation for blind queue workflows.
+
+    Issue #175 blocks a reviewer queue task from silently becoming author
+    implementation work. The workflow may use both namespaces only when that
+    mixed use is explicit, justified, and non-mutating; author mutations after
+    a reviewer queue task require an explicit operator authorization.
+
+    *proof* keys:
+    ``task_role`` ('reviewer' or 'author'), ``task_kind`` (for example
+    'blind_pr_queue_review'), ``reviewer_namespace_used``,
+    ``author_namespace_used``, ``author_mutations`` (list), ``review_mutations``
+    (list), ``operator_authorized_author_work``, ``mixed_namespace_justification``,
+    ``scratch_evidence_claimed``, and ``scratch_evidence_durable``.
+    """
+    proof = proof or {}
+    task_role = (proof.get("task_role") or "").strip().lower()
+    task_kind = (proof.get("task_kind") or "").strip().lower()
+    author_mutations = list(proof.get("author_mutations") or [])
+    review_mutations = list(proof.get("review_mutations") or [])
+    reviewer_used = bool(proof.get("reviewer_namespace_used"))
+    author_used = bool(proof.get("author_namespace_used"))
+    authorized = bool(proof.get("operator_authorized_author_work"))
+    mixed_justification = (
+        proof.get("mixed_namespace_justification") or ""
+    ).strip()
+    scratch_claimed = bool(proof.get("scratch_evidence_claimed"))
+    scratch_durable = bool(proof.get("scratch_evidence_durable"))
+
+    reasons = []
+    violations = []
+
+    if task_role not in {"reviewer", "author"}:
+        reasons.append("task role missing or unknown; role boundary unproven")
+
+    if task_role == "reviewer":
+        if author_mutations and not authorized:
+            violations.append(
+                "reviewer task performed author mutations without explicit "
+                "operator authorization"
+            )
+        if author_used and not mixed_justification:
+            reasons.append(
+                "reviewer task used author namespace without an explicit "
+                "justification"
+            )
+        if task_kind == "blind_pr_queue_review" and author_mutations:
+            if not authorized:
+                violations.append(
+                    "blind PR queue review silently pivoted into author "
+                    "implementation"
+                )
+    elif task_role == "author":
+        if review_mutations:
+            violations.append(
+                "author task performed reviewer-only mutations"
+            )
+
+    if reviewer_used and author_used and not mixed_justification:
+        reasons.append(
+            "mixed reviewer+author namespace use was not reported as a "
+            "role-boundary event"
+        )
+
+    if scratch_claimed and not scratch_durable:
+        reasons.append(
+            "scratch-only notes were claimed as durable evidence"
+        )
+
+    if violations:
+        status = "violation"
+        safe_next_action = "stop; report role-boundary violation"
+    elif reasons:
+        status = "warning"
+        safe_next_action = "downgrade final report; do not claim A-level proof"
+    else:
+        status = "clean"
+        safe_next_action = "proceed"
+
+    return {
+        "status": status,
+        "clean": status == "clean",
+        "reasons": reasons,
+        "violations": violations,
+        "safe_next_action": safe_next_action,
+    }
+
+
 def build_final_report(checkout_proof, inventory, validation, contamination,
                        identity_eligible, merge_performed,
-                       issue_status_verified):
+                       issue_status_verified, role_boundary=None):
     """Required behavior 6 + acceptance criteria: one report, distinct proofs.
 
     Combines the individual proof verdicts into the final-report fields the
@@ -348,6 +436,12 @@ def build_final_report(checkout_proof, inventory, validation, contamination,
     checkout_proven = bool(checkout_proof.get("proven"))
     validation_claimable = bool(validation.get("claimable"))
     validation_strong = validation.get("verdict") == "strong"
+    role_boundary = role_boundary or {
+        "status": "warning",
+        "reasons": ["role-boundary proof missing"],
+        "violations": [],
+    }
+    role_status = role_boundary.get("status", "warning")
 
     downgrade_reasons = []
     if not identity_eligible:
@@ -367,6 +461,9 @@ def build_final_report(checkout_proof, inventory, validation, contamination,
         downgrade_reasons.append(
             f"session contamination status is '{contamination_status}'"
         )
+    if role_status != "clean":
+        downgrade_reasons.append(f"role boundary status is '{role_status}'")
+        downgrade_reasons.extend(role_boundary.get("reasons", []))
     if not issue_status_verified:
         downgrade_reasons.append("linked issue status not verified")
 
@@ -374,6 +471,7 @@ def build_final_report(checkout_proof, inventory, validation, contamination,
         identity_eligible
         and checkout_proven
         and contamination_status == "clean"
+        and role_status == "clean"
         and validation_claimable
         and validation.get("verdict") != "invalid"
     )
@@ -384,6 +482,7 @@ def build_final_report(checkout_proof, inventory, validation, contamination,
             "merge was performed/claimed although the proofs did not allow "
             "one; this run is blocked, not graded"
         )
+    violations.extend(role_boundary.get("violations", []))
 
     if violations:
         grade = "blocked"
@@ -400,6 +499,7 @@ def build_final_report(checkout_proof, inventory, validation, contamination,
         "pr_author_distinct_from_reviewer":
             contamination_status in ("clean",),
         "session_contamination": contamination_status,
+        "role_boundary": role_status,
         "inventory_complete": bool(inventory.get("complete")),
         "validated_on_pinned_head": checkout_proven and validation_claimable,
         "validation_passed":

skills/llm-project-workflow/SKILL.md

@@ -217,19 +217,27 @@ Worktree folder = branch with `/` replaced by `-`
    Both configured repos must be reported with state filter, pagination proof,
    and open-PR count (`review_proofs.assess_inventory_completeness` and
    `resolve_repos_from_user_reference`).
-7. Inspect the full diff; confirm scope matches the linked issue; flag unrelated files.
-8. Run the tests. Validation reporting must include the exact command and
+7. **Role-boundary proof (#175):** a reviewer queue task must not silently
+   become author implementation. If no eligible PR exists, stop with the
+   queue report. Do not claim issues, create branches, commit, push, or open
+   PRs unless the operator explicitly retasks the run as author work. Mixed
+   reviewer+author namespace use must be reported with a justification, and
+   scratch-only notes are not durable evidence unless posted or committed
+   intentionally (`review_proofs.assess_role_boundary`).
+8. Inspect the full diff; confirm scope matches the linked issue; flag unrelated files.
+9. Run the tests. Validation reporting must include the exact command and
    exact results: pass/fail, counts of tests passed/skipped/failed, any
    ignored paths and why they are safe to ignore, and whether the command
    differs from the repository's canonical validation command. Only claim a
    validation result after the command has completed and its output has
    been read (`review_proofs.assess_validation_report`).
-9. **Do not merge if checks fail. Do not merge if the reviewer is the author.**
-10. The final report must distinguish (`review_proofs.build_final_report`):
+10. **Do not merge if checks fail. Do not merge if the reviewer is the author.**
+11. The final report must distinguish (`review_proofs.build_final_report`):
     identity eligible; PR author different from reviewer; session
-    contamination absent (with evidence); validation performed on the pinned
-    head; merge performed; issue status verified. If any proof is missing,
-    stop or downgrade the result instead of merging confidently.
+    contamination absent (with evidence); role boundary clean; validation
+    performed on the pinned head; merge performed; issue status verified. If
+    any proof is missing, stop or downgrade the result instead of merging
+    confidently.
 
 ## G. Merge / cleanup workflow
 

skills/llm-project-workflow/templates/review-pr.md

@@ -23,6 +23,10 @@ Rules (llm-project-workflow):
 - You must NOT be the PR author. If the authenticated user == PR author, stop.
   A different LLM-Agent-SHA does NOT make you a different actor — only a
   different authenticated Gitea user does (docs/llm-agent-sha.md).
+- Do not pivot from a reviewer queue task into author implementation unless
+  the operator explicitly retasks the run. If author namespace was used, the
+  final report must justify why; author mutations after reviewer queue work
+  without explicit authorization are a role-boundary violation.
 - Do not merge if any check fails.
 
 Steps:
@@ -39,6 +43,10 @@ Steps:
    cannot evidence whether this session authored/touched the PR branch,
    report contamination as UNKNOWN (not contaminated, not clean) and choose
    another PR or stop.
+   Role-boundary claims must also be evidence-backed (#175): report whether
+   reviewer namespace, author namespace, author mutations, or review mutations
+   occurred. Use `review_proofs.assess_role_boundary`; if it is not clean,
+   downgrade or stop instead of claiming an A-level run.
 5. scripts/worktree-review <pr-head-branch>          # detached, branches/review-*
    cd branches/review-<pr-head-branch-slug>
 6. Checkout proof (#173) — prove and state, before any diff review or

tests/test_review_proofs.py

@@ -23,6 +23,7 @@ sys.path.insert(0, str(__import__("pathlib").Path(__file__).resolve().parent.par
 from review_proofs import (  # noqa: E402
     assess_controller_handoff,
     assess_inventory_completeness,
+    assess_role_boundary,
     assess_self_review_contamination,
     assess_validation_report,
     build_final_report,
@@ -99,6 +100,21 @@ def _good_contamination():
     )
 
 
+def _good_role_boundary():
+    return assess_role_boundary(
+        {
+            "task_role": "reviewer",
+            "task_kind": "blind_pr_queue_review",
+            "reviewer_namespace_used": True,
+            "author_namespace_used": False,
+            "author_mutations": [],
+            "review_mutations": [],
+            "operator_authorized_author_work": False,
+            "scratch_evidence_claimed": False,
+        }
+    )
+
+
 class TestCheckoutProof(unittest.TestCase):
     """Required behavior 1 + 2: prove HEAD == pinned PR head or stop."""
 
@@ -368,6 +384,74 @@ class TestSelfReviewContamination(unittest.TestCase):
         self.assertEqual(result["status"], "unknown")
 
 
+class TestRoleBoundary(unittest.TestCase):
+    """Issue #175: reviewer queue tasks must not pivot into author work."""
+
+    def test_reviewer_queue_without_author_mutations_is_clean(self):
+        result = _good_role_boundary()
+        self.assertEqual(result["status"], "clean")
+        self.assertEqual(result["violations"], [])
+
+    def test_reviewer_queue_author_mutation_without_authorization_violates(self):
+        result = assess_role_boundary(
+            {
+                "task_role": "reviewer",
+                "task_kind": "blind_pr_queue_review",
+                "reviewer_namespace_used": True,
+                "author_namespace_used": True,
+                "author_mutations": ["claim issue #171", "push branch"],
+                "operator_authorized_author_work": False,
+                "mixed_namespace_justification": (
+                    "author namespace was used for implementation"
+                ),
+            }
+        )
+        self.assertEqual(result["status"], "violation")
+        self.assertTrue(any("pivot" in r for r in result["violations"]))
+
+    def test_mixed_namespace_use_without_justification_is_warning(self):
+        result = assess_role_boundary(
+            {
+                "task_role": "reviewer",
+                "task_kind": "blind_pr_queue_review",
+                "reviewer_namespace_used": True,
+                "author_namespace_used": True,
+                "author_mutations": [],
+            }
+        )
+        self.assertEqual(result["status"], "warning")
+        self.assertTrue(
+            any("mixed" in r.lower() for r in result["reasons"])
+        )
+
+    def test_author_task_cannot_perform_review_mutations(self):
+        result = assess_role_boundary(
+            {
+                "task_role": "author",
+                "reviewer_namespace_used": False,
+                "author_namespace_used": True,
+                "review_mutations": ["approve PR"],
+            }
+        )
+        self.assertEqual(result["status"], "violation")
+        self.assertTrue(
+            any("reviewer-only" in r for r in result["violations"])
+        )
+
+    def test_scratch_only_notes_are_not_durable_evidence(self):
+        result = assess_role_boundary(
+            {
+                "task_role": "reviewer",
+                "task_kind": "blind_pr_queue_review",
+                "reviewer_namespace_used": True,
+                "scratch_evidence_claimed": True,
+                "scratch_evidence_durable": False,
+            }
+        )
+        self.assertEqual(result["status"], "warning")
+        self.assertTrue(any("scratch-only" in r for r in result["reasons"]))
+
+
 class TestFinalReport(unittest.TestCase):
     """Required behavior 6 + acceptance criteria: the report must
     distinguish each proof, and only a fully proven run earns an "A"."""
@@ -381,6 +465,7 @@ class TestFinalReport(unittest.TestCase):
             "identity_eligible": True,
             "merge_performed": False,
             "issue_status_verified": True,
+            "role_boundary": _good_role_boundary(),
         }
         kwargs.update(overrides)
         return build_final_report(**kwargs)
@@ -393,6 +478,7 @@ class TestFinalReport(unittest.TestCase):
         self.assertTrue(report["identity_eligible"])
         self.assertTrue(report["pr_author_distinct_from_reviewer"])
         self.assertEqual(report["session_contamination"], "clean")
+        self.assertEqual(report["role_boundary"], "clean")
         self.assertTrue(report["validated_on_pinned_head"])
         self.assertFalse(report["merge_performed"])
         self.assertTrue(report["issue_status_verified"])
@@ -465,6 +551,37 @@ class TestFinalReport(unittest.TestCase):
         self.assertNotEqual(report["grade"], "A")
         self.assertFalse(report["merge_allowed"])
 
+    def test_missing_role_boundary_downgrades_and_blocks_merge(self):
+        kwargs = {
+            "checkout_proof": _good_checkout(),
+            "inventory": _good_inventory(),
+            "validation": _good_validation(),
+            "contamination": _good_contamination(),
+            "identity_eligible": True,
+            "merge_performed": False,
+            "issue_status_verified": True,
+        }
+        report = build_final_report(**kwargs)
+        self.assertNotEqual(report["grade"], "A")
+        self.assertFalse(report["merge_allowed"])
+        self.assertEqual(report["role_boundary"], "warning")
+
+    def test_role_boundary_violation_blocks_report(self):
+        boundary = assess_role_boundary(
+            {
+                "task_role": "reviewer",
+                "task_kind": "blind_pr_queue_review",
+                "reviewer_namespace_used": True,
+                "author_namespace_used": True,
+                "author_mutations": ["create PR"],
+                "operator_authorized_author_work": False,
+                "mixed_namespace_justification": "implementation pivot",
+            }
+        )
+        report = self._report(role_boundary=boundary)
+        self.assertEqual(report["grade"], "blocked")
+        self.assertFalse(report["merge_allowed"])
+
 
 class TestStdoutIsolation(unittest.TestCase):
     """Regression test for #178: tests must not close or corrupt stdout/stderr