Merge pull request 'feat: composable final-report validator for reviewer and reconciliation handoffs (Closes #327)' (#348) from feat/issue-327-final-report-validator-rules into master

final_report_validator.py

@@ -0,0 +1,878 @@
+"""Composable final-report validator for reviewer and reconciliation handoffs (#327).
+
+Single entry point ``assess_final_report_validator`` runs task-specific rule
+modules and returns structured findings suitable for controller handoff
+``Blocker`` / ``Safe next action`` fields.
+"""
+
+from __future__ import annotations
+
+import inspect
+import re
+from typing import Any, Callable
+
+from review_proofs import (
+    HANDOFF_HEADING,
+    assess_controller_handoff,
+    assess_email_disclosure,
+    assess_issue_filing_final_report,
+    assess_mutation_ledger_report,
+    assess_review_mutation_final_report,
+    assess_validation_report,
+)
+
+FINAL_REPORT_TASK_KINDS = frozenset({
+    "review_pr",
+    "reconcile_already_landed",
+    "author_issue",
+    "work_issue",
+    "issue_filing",
+    "issue_selection",
+    "inventory",
+})
+
+_TASK_KIND_ALIASES = {
+    "review": "review_pr",
+    "review-merge-pr": "review_pr",
+    "reconcile-landed-pr": "reconcile_already_landed",
+    "reconcile_already_landed": "reconcile_already_landed",
+    "work-issue": "work_issue",
+    "create_issue": "issue_filing",
+}
+
+_HANDOFF_ROLE_BY_TASK = {
+    "review_pr": "review",
+    "reconcile_already_landed": None,
+    "author_issue": "author",
+    "work_issue": "author",
+    "issue_filing": "issue_filing",
+    "issue_selection": None,
+    "inventory": "inventory",
+}
+
+_LEGACY_WORKSPACE_MUTATIONS_RE = re.compile(
+    r"^\s*[-*]?\s*workspace\s+mutations\s*:",
+    re.IGNORECASE | re.MULTILINE,
+)
+_VAGUE_MUTATIONS_NONE_RE = re.compile(
+    r"^\s*[-*]?\s*mutations\s*:\s*none\s*$",
+    re.IGNORECASE | re.MULTILINE,
+)
+_BARE_PYTEST_RE = re.compile(r"\b(?:pytest|python\s+-m\s+pytest)\b", re.IGNORECASE)
+_VALIDATION_ENV_RE = re.compile(
+    r"(?:working directory|cwd|executed (?:in|from)|command\s+path)\s*:|\bin\s+branches/",
+    re.IGNORECASE,
+)
+_MAIN_CHECKOUT_RE = re.compile(
+    r"(?:/Gitea-Tools(?:/|$)(?!branches/)|main checkout|shared checkout|control checkout)",
+    re.IGNORECASE,
+)
+_SAME_AS_MASTER_RE = re.compile(
+    r"same as master|same failures as master|matches master baseline",
+    re.IGNORECASE,
+)
+_BASELINE_WORKTREE_RE = re.compile(
+    r"baseline worktree",
+    re.IGNORECASE,
+)
+_APPROVAL_CLAIM_RE = re.compile(
+    r"(?:submitted\s+['\"]approve['\"]|review decision\s*:\s*approve|claims?\s+approve)",
+    re.IGNORECASE,
+)
+_FULL_SUITE_FAIL_RE = re.compile(
+    r"(?:full suite failed|full test suite failed|\d+\s+failed(?:,|\s))",
+    re.IGNORECASE,
+)
+_ALREADY_LANDED_RE = re.compile(
+    r"already[- ]landed|ancestor proof\s*:\s*passed|eligibility class\s*:\s*already_landed",
+    re.IGNORECASE,
+)
+_ELIGIBLE_REVIEW_RE = re.compile(
+    r"eligible for (?:review|merge)|ready for (?:review|merge)|next eligible pr",
+    re.IGNORECASE,
+)
+_REVIEWED_LANDED_RE = re.compile(
+    r"(?:reviewed|approved).{0,40}already[- ]landed|already[- ]landed.{0,40}(?:reviewed|eligible)",
+    re.IGNORECASE | re.DOTALL,
+)
+_RECONCILE_STALE_FIELDS = (
+    "pr number opened",
+    "pinned reviewed head",
+    "scratch worktree used",
+    "review decision",
+    "merge result",
+)
+_RECONCILE_REQUIRED_FIELDS = (
+    "Selected PR",
+    "Eligibility class",
+    "Linked issue live status",
+    "Safe next action",
+    "No review/merge confirmation",
+)
+_LINKED_ISSUE_STATUS_RE = re.compile(
+    r"linked issue(?:\s+live)?\s+status\s*:\s*(.+)$",
+    re.IGNORECASE | re.MULTILINE,
+)
+_LINKED_ISSUE_NUMBER_RE = re.compile(
+    r"linked issue\s*:\s*#?(\d+)",
+    re.IGNORECASE,
+)
+_INVENTORY_COMPLETE_RE = re.compile(
+    r"inventory\s+(?:complete|completeness\s*:\s*complete)",
+    re.IGNORECASE,
+)
+_PAGINATION_PROOF_RE = re.compile(
+    r"(?:pagination\s+complete|final page|no next page|total\s+(?:open\s+)?pr\s+count|traversed\s+all\s+pages)",
+    re.IGNORECASE,
+)
+_GIT_FETCH_RE = re.compile(r"\bgit\s+fetch\b", re.IGNORECASE)
+_READONLY_DIAG_RE = re.compile(
+    r"read[- ]only diagnostics\s*:\s*(.+)$",
+    re.IGNORECASE | re.MULTILINE,
+)
+_MUTATION_CATEGORY_FIELDS = (
+    "file edits by reviewer",
+    "worktree/index mutations",
+    "git ref mutations",
+    "mcp/gitea mutations",
+    "reconciliation mutations",
+    "file edits by reconciler",
+)
+
+
+def _normalize_task_kind(task_kind: str | None) -> str:
+    raw = (task_kind or "").strip().lower().replace(" ", "_")
+    return _TASK_KIND_ALIASES.get(raw, raw)
+
+
+def validator_finding(
+    rule_id: str,
+    severity: str,
+    field: str,
+    reason: str,
+    safe_next_action: str,
+) -> dict[str, str]:
+    """Structured finding for controller handoff consumption."""
+    return {
+        "rule_id": rule_id,
+        "severity": severity,
+        "field": field,
+        "reason": reason,
+        "safe_next_action": safe_next_action,
+    }
+
+
+def _findings_from_reasons(
+    rule_id: str,
+    reasons: list[str],
+    *,
+    field: str = "",
+    severity: str = "downgrade",
+    safe_next_action: str = "downgrade final report; repair missing proof fields",
+) -> list[dict[str, str]]:
+    return [
+        validator_finding(rule_id, severity, field, reason, safe_next_action)
+        for reason in reasons
+    ]
+
+
+def _handoff_fields(report_text: str) -> dict[str, str]:
+    from review_proofs import _handoff_section_lines  # local import avoids cycle at load
+
+    section = _handoff_section_lines(report_text) or []
+    fields: dict[str, str] = {}
+    for line in section:
+        stripped = line.strip().lstrip("-*").strip()
+        if ":" not in stripped:
+            continue
+        key, value = stripped.split(":", 1)
+        fields[key.strip().lower()] = value.strip()
+    return fields
+
+
+def _rule_shared_controller_handoff(
+    report_text: str,
+    task_kind: str,
+    *,
+    local_edits: bool = False,
+) -> list[dict[str, str]]:
+    role = _HANDOFF_ROLE_BY_TASK.get(task_kind)
+    result = assess_controller_handoff(
+        report_text,
+        role=role,
+        local_edits=local_edits,
+    )
+    verdict = result.get("verdict")
+    if verdict == "complete":
+        return []
+    severity = "block" if verdict == "missing" else "downgrade"
+    return _findings_from_reasons(
+        "shared.controller_handoff",
+        result.get("reasons") or [],
+        field="Controller Handoff",
+        severity=severity,
+        safe_next_action=(
+            "add a complete Controller Handoff section with task-appropriate fields"
+            if verdict == "missing"
+            else "fill missing controller handoff fields before submission"
+        ),
+    )
+
+
+def _rule_shared_email_disclosure(report_text: str) -> list[dict[str, str]]:
+    result = assess_email_disclosure(report_text)
+    if result.get("proven"):
+        return []
+    return _findings_from_reasons(
+        "shared.email_disclosure",
+        result.get("reasons") or [],
+        field="Identity",
+        severity="downgrade",
+        safe_next_action="use username / profile identity; remove personal email",
+    )
+
+
+def _rule_reviewer_legacy_workspace_mutations(
+    report_text: str,
+    *,
+    mutations_observed: bool = False,
+) -> list[dict[str, str]]:
+    if not mutations_observed:
+        return []
+    if not _LEGACY_WORKSPACE_MUTATIONS_RE.search(report_text or ""):
+        return []
+    return [
+        validator_finding(
+            "reviewer.legacy_workspace_mutations",
+            "block",
+            "Workspace mutations",
+            "legacy 'Workspace mutations' field used after observed mutations; "
+            "use precise mutation categories instead",
+            "replace Workspace mutations with file/worktree/git/MCP category fields",
+        )
+    ]
+
+
+def _rule_reviewer_vague_mutations_none(
+    report_text: str,
+    *,
+    action_log: list[dict] | None = None,
+    mutations_observed: bool = False,
+) -> list[dict[str, str]]:
+    performed = any(
+        e.get("performed") is not False and not e.get("gated_rejected")
+        for e in (action_log or [])
+    )
+    if not (mutations_observed or performed):
+        return []
+    if not _VAGUE_MUTATIONS_NONE_RE.search(report_text or ""):
+        return []
+    return [
+        validator_finding(
+            "reviewer.vague_mutations_none",
+            "block",
+            "Mutations",
+            "report claims 'Mutations: none' while mutations were observed",
+            "list performed mutations under precise category fields",
+        )
+    ]
+
+
+def _rule_reviewer_mutation_categories(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    lower = text.lower()
+    if "mutations:" not in lower and "mutation categories" not in lower:
+        return [
+            validator_finding(
+                "reviewer.mutation_categories",
+                "downgrade",
+                "Mutations",
+                "reviewer handoff lacks precise mutation category fields",
+                "report file edits, worktree/index, git ref, MCP/Gitea, and review mutations separately",
+            )
+        ]
+    if any(term in lower for term in ("workspace mutations", "mutations: none")):
+        return []
+    if any(field in lower for field in _MUTATION_CATEGORY_FIELDS):
+        return []
+    vague_only = "mutations:" in lower and not any(
+        marker in lower
+        for marker in ("file edits", "worktree", "git ref", "mcp/gitea", "review mutation")
+    )
+    if vague_only:
+        return [
+            validator_finding(
+                "reviewer.mutation_categories",
+                "downgrade",
+                "Mutations",
+                "mutation section is vague; precise categories are required",
+                "split mutations into file edits, worktree/index, git ref, MCP/Gitea, review",
+            )
+        ]
+    return []
+
+
+def _rule_reviewer_git_fetch_readonly(
+    report_text: str,
+    *,
+    action_log: list[dict] | None = None,
+) -> list[dict[str, str]]:
+    text = report_text or ""
+    fetch_observed = any(
+        _GIT_FETCH_RE.search(str(e.get("command") or e.get("action") or ""))
+        for e in (action_log or [])
+    ) or _GIT_FETCH_RE.search(text)
+    if not fetch_observed:
+        return []
+
+    readonly_match = _READONLY_DIAG_RE.search(text)
+    readonly_value = (readonly_match.group(1) if readonly_match else "").strip().lower()
+    fields = _handoff_fields(text)
+    git_ref_value = fields.get("git ref mutations", "").strip().lower()
+    git_ref_reported = bool(git_ref_value) and git_ref_value not in {
+        "none",
+        "n/a",
+        "no",
+    }
+    if git_ref_reported:
+        return []
+    if readonly_match and _GIT_FETCH_RE.search(readonly_value):
+        return [
+            validator_finding(
+                "reviewer.git_fetch_readonly",
+                "block",
+                "Git ref mutations",
+                "git fetch occurred but was classified only as read-only diagnostics",
+                "classify git fetch under Git ref mutations, not read-only diagnostics",
+            )
+        ]
+    if not git_ref_reported and _GIT_FETCH_RE.search(text):
+        return [
+            validator_finding(
+                "reviewer.git_fetch_readonly",
+                "downgrade",
+                "Git ref mutations",
+                "git fetch mentioned without Git ref mutation classification",
+                "record git fetch under Git ref mutations",
+            )
+        ]
+    return []
+
+
+def _rule_reviewer_validation_command(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not _BARE_PYTEST_RE.search(text):
+        return []
+    if _VALIDATION_ENV_RE.search(text):
+        return []
+    return [
+        validator_finding(
+            "reviewer.validation_command_proof",
+            "downgrade",
+            "Validation",
+            "bare pytest command without working-directory or executable path proof",
+            "report exact validation command, cwd, and path proof for bare commands",
+        )
+    ]
+
+
+def _rule_reviewer_linked_issue(
+    report_text: str,
+    *,
+    linked_issue_lock: dict | None = None,
+) -> list[dict[str, str]]:
+    lock = linked_issue_lock or {}
+    expected_issue = lock.get("issue_number")
+    expected_pr = lock.get("pr_number")
+    if expected_issue is None and expected_pr is None:
+        return []
+
+    text = report_text or ""
+    findings: list[dict[str, str]] = []
+    reported_issue = None
+    issue_match = _LINKED_ISSUE_NUMBER_RE.search(text)
+    if issue_match:
+        reported_issue = int(issue_match.group(1))
+    elif expected_issue is not None:
+        if f"#{expected_issue}" not in text and f"issue {expected_issue}" not in text.lower():
+            findings.append(
+                validator_finding(
+                    "reviewer.linked_issue_missing",
+                    "downgrade",
+                    "Linked issue",
+                    f"linked issue #{expected_issue} not documented in final report",
+                    "document the linked issue number matching the selected PR",
+                )
+            )
+
+    if expected_issue is not None and reported_issue is not None and reported_issue != expected_issue:
+        findings.append(
+            validator_finding(
+                "reviewer.linked_issue_mismatch",
+                "block",
+                "Linked issue",
+                f"reported linked issue #{reported_issue} does not match "
+                f"selected PR linked issue #{expected_issue}",
+                "reconcile linked issue proof with the selected PR before reporting status",
+            )
+        )
+
+    if expected_pr is not None:
+        pr_tokens = (f"pr #{expected_pr}", f"pr{expected_pr}", f"#pr{expected_pr}")
+        if not any(token in text.lower().replace(" ", "") for token in pr_tokens):
+            if f"#{expected_pr}" not in text:
+                findings.append(
+                    validator_finding(
+                        "reviewer.linked_pr_missing",
+                        "downgrade",
+                        "Selected PR",
+                        f"selected PR #{expected_pr} not consistently referenced",
+                        "reference the selected PR number in handoff and diagnostics",
+                    )
+                )
+    return findings
+
+
+def _rule_reviewer_baseline_on_failure(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not (_APPROVAL_CLAIM_RE.search(text) and _FULL_SUITE_FAIL_RE.search(text)):
+        return []
+    if _BASELINE_WORKTREE_RE.search(text) and "baseline worktree path" in text.lower():
+        return []
+    return [
+        validator_finding(
+            "reviewer.baseline_on_full_suite_failure",
+            "block",
+            "Baseline worktree",
+            "approval claimed after full-suite failure without baseline worktree proof",
+            "create a clean branches/ baseline worktree and report path plus results",
+        )
+    ]
+
+
+def _rule_reviewer_main_checkout_baseline(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not _SAME_AS_MASTER_RE.search(text):
+        return []
+    if _BASELINE_WORKTREE_RE.search(text):
+        return []
+    return [
+        validator_finding(
+            "reviewer.main_checkout_baseline",
+            "block",
+            "Baseline worktree",
+            "vague 'same as master' claim without clean baseline worktree proof",
+            "validate against a clean branches/ baseline worktree; do not use main checkout",
+        )
+    ]
+
+
+def _rule_reviewer_main_checkout_path(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if "baseline worktree path" not in text.lower():
+        return []
+    fields = _handoff_fields(text)
+    baseline_path = ""
+    for key, value in fields.items():
+        if key.startswith("baseline worktree path"):
+            baseline_path = value
+            break
+    if baseline_path and _MAIN_CHECKOUT_RE.search(baseline_path):
+        return [
+            validator_finding(
+                "reviewer.main_checkout_path",
+                "block",
+                "Baseline worktree path",
+                "baseline validation used main/shared checkout instead of branches/ worktree",
+                "recreate baseline proof under branches/ and report that path",
+            )
+        ]
+    return []
+
+
+def _rule_reviewer_already_landed_eligible(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not _ALREADY_LANDED_RE.search(text):
+        return []
+    if not _ELIGIBLE_REVIEW_RE.search(text):
+        return []
+    return [
+        validator_finding(
+            "reviewer.already_landed_eligible_wording",
+            "block",
+            "Eligibility class",
+            "already-landed PR described as eligible for review or merge",
+            "classify as ALREADY_LANDED_RECONCILE_REQUIRED and stop normal review",
+        )
+    ]
+
+
+def _rule_reconcile_controller_handoff(report_text: str) -> list[dict[str, str]]:
+    from review_proofs import _handoff_section_lines
+
+    if _handoff_section_lines(report_text) is None:
+        return [
+            validator_finding(
+                "reconcile.controller_handoff",
+                "block",
+                "Controller Handoff",
+                f"final report has no section titled exactly '{HANDOFF_HEADING}'",
+                "add a reconciliation Controller Handoff section",
+            )
+        ]
+
+    fields = _handoff_fields(report_text)
+    missing = [
+        name
+        for name in _RECONCILE_REQUIRED_FIELDS
+        if not (fields.get(name.lower()) or "").strip()
+    ]
+    if missing:
+        return [
+            validator_finding(
+                "reconcile.controller_handoff",
+                "downgrade",
+                "Controller Handoff",
+                f"reconciliation handoff missing required field: {field}",
+                "complete the reconciliation handoff schema before submission",
+            )
+            for field in missing
+        ]
+    return []
+
+
+def _rule_reconcile_stale_author_fields(report_text: str) -> list[dict[str, str]]:
+    text = (report_text or "").lower()
+    findings = []
+    for field in _RECONCILE_STALE_FIELDS:
+        if f"{field}:" in text:
+            findings.append(
+                validator_finding(
+                    "reconcile.stale_author_reviewer_fields",
+                    "block",
+                    field,
+                    f"reconciliation handoff includes stale author/reviewer field '{field}'",
+                    "use reconciliation schema only; remove author/reviewer review fields",
+                )
+            )
+    return findings
+
+
+def _rule_reconcile_eligible_reviewed(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not _ALREADY_LANDED_RE.search(text):
+        return []
+    findings = []
+    if _ELIGIBLE_REVIEW_RE.search(text):
+        findings.append(
+            validator_finding(
+                "reconcile.eligible_reviewed_wording",
+                "block",
+                "Eligibility class",
+                "already-landed PR described as eligible for review or merge",
+                "report ALREADY_LANDED_RECONCILE_REQUIRED and reconciliation-only next steps",
+            )
+        )
+    if _REVIEWED_LANDED_RE.search(text):
+        findings.append(
+            validator_finding(
+                "reconcile.reviewed_landed_wording",
+                "block",
+                "Current status",
+                "already-landed PR described as reviewed or approval-eligible",
+                "use reconciliation status wording only",
+            )
+        )
+    return findings
+
+
+def _rule_reconcile_linked_issue_live(
+    report_text: str,
+    *,
+    linked_issue_lock: dict | None = None,
+) -> list[dict[str, str]]:
+    lock = linked_issue_lock or {}
+    text = report_text or ""
+    status_match = _LINKED_ISSUE_STATUS_RE.search(text)
+    if not status_match:
+        return []
+
+    status_value = status_match.group(1).strip().lower()
+    if "not verified" in status_value:
+        return []
+
+    live_proof = lock.get("live_fetched") is True or lock.get("verified_live") is True
+    if live_proof:
+        return []
+
+    if status_value not in {"", "none", "n/a", "unknown"}:
+        return [
+            validator_finding(
+                "reconcile.linked_issue_live_status",
+                "downgrade",
+                "Linked issue live status",
+                "linked issue status reported without live verification proof in this session",
+                "fetch linked issue live or report 'not verified in this session'",
+            )
+        ]
+    return []
+
+
+def _rule_reconcile_pagination_proof(report_text: str) -> list[dict[str, str]]:
+    text = report_text or ""
+    if not _INVENTORY_COMPLETE_RE.search(text):
+        return []
+    if _PAGINATION_PROOF_RE.search(text):
+        return []
+    return [
+        validator_finding(
+            "reconcile.inventory_pagination_proof",
+            "downgrade",
+            "PR inventory",
+            "inventory completeness claimed without pagination or final-page proof",
+            "include pagination complete, final page, or total open PR count evidence",
+        )
+    ]
+
+
+def _rule_reviewer_validation_structured(
+    validation_report: dict | None,
+) -> list[dict[str, str]]:
+    if not validation_report:
+        return []
+    result = assess_validation_report(validation_report)
+    if result.get("verdict") == "strong":
+        return []
+    severity = "block" if result.get("verdict") == "invalid" else "downgrade"
+    return _findings_from_reasons(
+        "reviewer.validation_structured",
+        result.get("reasons") or [],
+        field="Validation",
+        severity=severity,
+        safe_next_action="provide exact validation command, output read proof, and counts",
+    )
+
+
+def _rule_reviewer_mutation_ledger(
+    report_text: str,
+    *,
+    action_log: list[dict] | None = None,
+) -> list[dict[str, str]]:
+    if not action_log:
+        return []
+    result = assess_mutation_ledger_report(report_text, action_log=action_log)
+    if result.get("proven"):
+        return []
+    return _findings_from_reasons(
+        "reviewer.mutation_ledger",
+        result.get("reasons") or [],
+        field="File edits by reviewer",
+        severity="block",
+        safe_next_action="align mutation ledger with observed file edits",
+    )
+
+
+def _rule_reviewer_review_mutation(
+    report_text: str,
+    *,
+    review_decision_lock: dict | None = None,
+) -> list[dict[str, str]]:
+    if not review_decision_lock:
+        return []
+    result = assess_review_mutation_final_report(report_text, review_decision_lock)
+    if result.get("complete"):
+        return []
+    return _findings_from_reasons(
+        "reviewer.review_mutation_proof",
+        result.get("reasons") or [],
+        field="Review mutations",
+        severity="downgrade",
+        safe_next_action="document exactly one live review mutation or authorized correction flow",
+    )
+
+
+_RULES_BY_TASK: dict[str, list[Callable[..., list[dict[str, str]]]]] = {
+    "review_pr": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+        _rule_reviewer_legacy_workspace_mutations,
+        _rule_reviewer_vague_mutations_none,
+        _rule_reviewer_mutation_categories,
+        _rule_reviewer_git_fetch_readonly,
+        _rule_reviewer_validation_command,
+        _rule_reviewer_validation_structured,
+        _rule_reviewer_linked_issue,
+        _rule_reviewer_baseline_on_failure,
+        _rule_reviewer_main_checkout_baseline,
+        _rule_reviewer_main_checkout_path,
+        _rule_reviewer_already_landed_eligible,
+        _rule_reviewer_mutation_ledger,
+        _rule_reviewer_review_mutation,
+    ],
+    "reconcile_already_landed": [
+        _rule_reconcile_controller_handoff,
+        _rule_shared_email_disclosure,
+        _rule_reconcile_stale_author_fields,
+        _rule_reconcile_eligible_reviewed,
+        _rule_reconcile_linked_issue_live,
+        _rule_reconcile_pagination_proof,
+        _rule_reviewer_git_fetch_readonly,
+        _rule_reviewer_legacy_workspace_mutations,
+        _rule_reviewer_vague_mutations_none,
+    ],
+    "author_issue": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+        _rule_reviewer_vague_mutations_none,
+    ],
+    "work_issue": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+        _rule_reviewer_vague_mutations_none,
+    ],
+    "issue_filing": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+    ],
+    "inventory": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+        _rule_reconcile_pagination_proof,
+    ],
+    "issue_selection": [
+        _rule_shared_controller_handoff,
+        _rule_shared_email_disclosure,
+    ],
+}
+
+
+def _aggregate_grade(findings: list[dict[str, str]]) -> tuple[str, bool, bool]:
+    blocked = any(f["severity"] == "block" for f in findings)
+    downgraded = any(f["severity"] == "downgrade" for f in findings)
+    if blocked:
+        return "blocked", True, downgraded
+    if downgraded:
+        return "downgraded", False, True
+    return "A", False, False
+
+
+def _primary_safe_next_action(findings: list[dict[str, str]]) -> str:
+    for severity in ("block", "downgrade", "warning"):
+        for finding in findings:
+            if finding["severity"] == severity:
+                return finding["safe_next_action"]
+    return "proceed"
+
+
+def _call_rule(
+    rule: Callable[..., list[dict[str, str]]],
+    report_text: str,
+    task_kind: str,
+    rule_kwargs: dict[str, Any],
+) -> list[dict[str, str]]:
+    if rule is _rule_shared_controller_handoff:
+        return rule(
+            report_text,
+            task_kind,
+            local_edits=bool(rule_kwargs.get("local_edits")),
+        )
+    if rule is _rule_reviewer_validation_structured:
+        return rule(rule_kwargs.get("validation_report"))
+    if rule is _rule_reconcile_controller_handoff:
+        return rule(report_text)
+
+    bound: dict[str, Any] = {}
+    for name, param in inspect.signature(rule).parameters.items():
+        if name == "report_text":
+            bound[name] = report_text
+        elif name in rule_kwargs:
+            bound[name] = rule_kwargs[name]
+        elif param.default is inspect.Parameter.empty and param.kind != inspect.Parameter.VAR_KEYWORD:
+            continue
+    return rule(**bound)
+
+
+def assess_final_report_validator(
+    report_text: str,
+    task_kind: str,
+    *,
+    review_decision_lock: dict | None = None,
+    linked_issue_lock: dict | None = None,
+    validation_report: dict | None = None,
+    action_log: list[dict] | None = None,
+    mutations_observed: bool = False,
+    local_edits: bool = False,
+    issue_filing_lock: dict | None = None,
+) -> dict[str, Any]:
+    """Validate final-report text against task-specific proof rules (#327).
+
+    Returns structured findings plus a composite grade suitable for workflow
+    controllers. Optional *proof_locks* carry live session facts used to fail
+    closed on stale or contradictory report fields.
+    """
+    normalized_kind = _normalize_task_kind(task_kind)
+    if normalized_kind not in FINAL_REPORT_TASK_KINDS:
+        return {
+            "task_kind": normalized_kind,
+            "grade": "blocked",
+            "blocked": True,
+            "downgraded": False,
+            "findings": [
+                validator_finding(
+                    "shared.unknown_task_kind",
+                    "block",
+                    "Task",
+                    f"unknown task kind '{task_kind}'",
+                    "use a supported task kind from FINAL_REPORT_TASK_KINDS",
+                )
+            ],
+            "checks": {},
+            "reasons": [f"unknown task kind '{task_kind}'"],
+            "safe_next_action": "set task_kind to a supported workflow mode",
+            "complete": False,
+        }
+
+    checks: dict[str, Any] = {}
+    findings: list[dict[str, str]] = []
+
+    if normalized_kind == "issue_filing" and issue_filing_lock is not None:
+        checks["issue_filing"] = assess_issue_filing_final_report(
+            report_text,
+            **issue_filing_lock,
+        )
+        if checks["issue_filing"].get("downgraded"):
+            findings.extend(
+                _findings_from_reasons(
+                    "issue_filing.composite",
+                    checks["issue_filing"].get("reasons") or [],
+                    field="Issue filing",
+                    severity="downgrade",
+                )
+            )
+
+    rule_kwargs = {
+        "review_decision_lock": review_decision_lock,
+        "linked_issue_lock": linked_issue_lock,
+        "validation_report": validation_report,
+        "action_log": action_log,
+        "mutations_observed": mutations_observed,
+        "local_edits": local_edits,
+    }
+
+    for rule in _RULES_BY_TASK.get(normalized_kind, ()):
+        findings.extend(_call_rule(rule, report_text, normalized_kind, rule_kwargs))
+
+    grade, blocked, downgraded = _aggregate_grade(findings)
+    reasons = [f"{f['rule_id']}: {f['reason']}" for f in findings]
+    return {
+        "task_kind": normalized_kind,
+        "grade": grade,
+        "blocked": blocked,
+        "downgraded": downgraded,
+        "findings": findings,
+        "checks": checks,
+        "reasons": reasons,
+        "safe_next_action": _primary_safe_next_action(findings),
+        "complete": grade == "A",
+        "handoff_heading": HANDOFF_HEADING,
+    }

review_proofs.py

@@ -3444,6 +3444,13 @@ def assess_email_disclosure(
     }
 
 
+def assess_final_report_validator(report_text, task_kind, **kwargs):
+    """#327: single entry point for task-specific final-report validation."""
+    from final_report_validator import assess_final_report_validator as _validate
+
+    return _validate(report_text, task_kind, **kwargs)
+
+
 _QUEUE_STATUS_REPORT_HINT = re.compile(
     r"queue[- ]status|selected pr:\s*none|no pr selected|"
     r"queue[- ]status[- ]only",

tests/test_final_report_validator.py

@@ -0,0 +1,273 @@
+"""Tests for composable final-report validator (#327)."""
+import sys
+import unittest
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from final_report_validator import (  # noqa: E402
+    FINAL_REPORT_TASK_KINDS,
+    assess_final_report_validator,
+    validator_finding,
+)
+
+
+def _review_handoff(**overrides):
+    lines = [
+        "## Controller Handoff",
+        "",
+        "- Task: review PR #203",
+        "- Repo: Scaled-Tech-Consulting/Gitea-Tools",
+        "- Role: reviewer",
+        "- Identity: sysadmin / prgs-reviewer",
+        "- Issue/PR: #182 / PR #203",
+        "- Branch/SHA: feat/x @ 0fdc8f582026b72a229d59a172c0a63ac4aaeaf9",
+        "- Files changed: review_proofs.py",
+        "- Validation: pytest tests/test_review_proofs.py -q in branches/review-203",
+        "- Mutations: review only",
+        "- Workspace mutations: none",
+        "- File edits by reviewer: none",
+        "- Worktree/index mutations: none",
+        "- Git ref mutations: git fetch prgs master",
+        "- MCP/Gitea mutations: review submitted",
+        "- Current status: PR open",
+        "- Blockers: none",
+        "- Next: await author",
+        "- Safety: no self-review; no self-merge",
+        "- Selected PR: #203",
+        "- Reviewer eligibility: eligible",
+        "- Pinned reviewed head: 0fdc8f582026b72a229d59a172c0a63ac4aaeaf9",
+        "- Worktree path: branches/review-203",
+        "- Worktree dirty: clean",
+        "- Scratch worktree used: yes (branches/review-203)",
+        "- Unrelated local mutations: none",
+        "- Review decision: request_changes",
+        "- Merge result: not attempted",
+        "- Linked issue: #182",
+        "- Linked issue status: open",
+        "- Cleanup status: none",
+    ]
+    text = "\n".join(lines)
+    for key, value in overrides.items():
+        text = text.replace(f"- {key}:", f"- {key}: {value}")
+    return text
+
+
+def _reconcile_handoff(**overrides):
+    lines = [
+        "## Controller Handoff",
+        "",
+        "- Task: reconcile already-landed PR #99",
+        "- Repo: Scaled-Tech-Consulting/Gitea-Tools",
+        "- Role: reconciler",
+        "- Identity: sysadmin / prgs-author",
+        "- Selected PR: #99",
+        "- Eligibility class: ALREADY_LANDED_RECONCILE_REQUIRED",
+        "- Linked issue: #98",
+        "- Linked issue live status: not verified in this session",
+        "- File edits by reconciler: none",
+        "- Git ref mutations: git fetch prgs master",
+        "- Reconciliation mutations: PR comment posted",
+        "- Current status: reconciliation complete",
+        "- Safe next action: none",
+        "- No review/merge confirmation: confirmed",
+    ]
+    text = "\n".join(lines)
+    for key, value in overrides.items():
+        if f"- {key}:" in text:
+            text = text.replace(f"- {key}:", f"- {key}: {value}")
+    return text
+
+
+class TestValidatorFinding(unittest.TestCase):
+    def test_finding_shape(self):
+        finding = validator_finding(
+            "reviewer.test",
+            "block",
+            "Validation",
+            "missing proof",
+            "repair validation proof",
+        )
+        self.assertEqual(finding["rule_id"], "reviewer.test")
+        self.assertEqual(finding["severity"], "block")
+        self.assertEqual(finding["field"], "Validation")
+
+
+class TestReviewPrRules(unittest.TestCase):
+    def test_clean_reviewer_report_passes(self):
+        result = assess_final_report_validator(
+            _review_handoff(),
+            "review_pr",
+            linked_issue_lock={"issue_number": 182, "pr_number": 203},
+        )
+        self.assertEqual(result["grade"], "A")
+        self.assertFalse(result["blocked"])
+        self.assertEqual(result["findings"], [])
+
+    def test_legacy_workspace_mutations_blocks(self):
+        report = _review_handoff() + "\n- Workspace mutations: none"
+        result = assess_final_report_validator(
+            report,
+            "review_pr",
+            mutations_observed=True,
+        )
+        self.assertEqual(result["grade"], "blocked")
+        rule_ids = [f["rule_id"] for f in result["findings"]]
+        self.assertIn("reviewer.legacy_workspace_mutations", rule_ids)
+
+    def test_vague_mutations_none_blocks(self):
+        report = _review_handoff().replace(
+            "- Mutations: review only",
+            "- Mutations: none",
+        )
+        result = assess_final_report_validator(
+            report,
+            "review_pr",
+            mutations_observed=True,
+        )
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(f["rule_id"] == "reviewer.vague_mutations_none" for f in result["findings"])
+        )
+
+    def test_bare_pytest_without_cwd_downgrades(self):
+        report = _review_handoff().replace(
+            "- Validation: pytest tests/test_review_proofs.py -q in branches/review-203",
+            "- Validation: pytest passed",
+        )
+        result = assess_final_report_validator(report, "review_pr")
+        self.assertEqual(result["grade"], "downgraded")
+        self.assertTrue(
+            any(f["rule_id"] == "reviewer.validation_command_proof" for f in result["findings"])
+        )
+
+    def test_linked_issue_mismatch_blocks(self):
+        report = _review_handoff().replace("- Linked issue: #182", "- Linked issue: #999")
+        result = assess_final_report_validator(
+            report,
+            "review_pr",
+            linked_issue_lock={"issue_number": 182, "pr_number": 203},
+        )
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(f["rule_id"] == "reviewer.linked_issue_mismatch" for f in result["findings"])
+        )
+
+    def test_approval_after_full_suite_failure_blocks_without_baseline(self):
+        report = (
+            _review_handoff()
+            .replace("- Review decision: request_changes", "- Review decision: approve")
+            + "\nSubmitted 'approve' after full suite failed with 3 failed tests."
+        )
+        result = assess_final_report_validator(report, "review_pr")
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(
+                f["rule_id"] == "reviewer.baseline_on_full_suite_failure"
+                for f in result["findings"]
+            )
+        )
+
+    def test_same_as_master_without_baseline_blocks(self):
+        report = _review_handoff() + "\nFailures are same as master baseline."
+        result = assess_final_report_validator(report, "review_pr")
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(f["rule_id"] == "reviewer.main_checkout_baseline" for f in result["findings"])
+        )
+
+    def test_already_landed_eligible_wording_blocks(self):
+        report = (
+            _review_handoff()
+            + "\nPR is already landed and eligible for review next."
+        )
+        result = assess_final_report_validator(report, "review_pr")
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(
+                f["rule_id"] == "reviewer.already_landed_eligible_wording"
+                for f in result["findings"]
+            )
+        )
+
+    def test_git_fetch_must_not_be_readonly_only(self):
+        report = (
+            _review_handoff()
+            .replace("- Git ref mutations: git fetch prgs master", "- Git ref mutations: none")
+            + "\n- Read-only diagnostics: git fetch prgs master"
+        )
+        result = assess_final_report_validator(
+            report,
+            "review_pr",
+            action_log=[{"command": "git fetch prgs master", "performed": True}],
+        )
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(f["rule_id"] == "reviewer.git_fetch_readonly" for f in result["findings"])
+        )
+
+
+class TestReconciliationRules(unittest.TestCase):
+    def test_clean_reconcile_report_passes(self):
+        result = assess_final_report_validator(
+            _reconcile_handoff(),
+            "reconcile_already_landed",
+        )
+        self.assertEqual(result["grade"], "A")
+
+    def test_stale_author_field_blocks(self):
+        report = _reconcile_handoff() + "\n- PR number opened: #12"
+        result = assess_final_report_validator(report, "reconcile_already_landed")
+        self.assertTrue(result["blocked"])
+        self.assertTrue(
+            any(
+                f["rule_id"] == "reconcile.stale_author_reviewer_fields"
+                for f in result["findings"]
+            )
+        )
+
+    def test_eligible_wording_for_landed_pr_blocks(self):
+        report = _reconcile_handoff() + "\nPR already landed and eligible for merge."
+        result = assess_final_report_validator(report, "reconcile_already_landed")
+        self.assertTrue(result["blocked"])
+
+    def test_linked_issue_status_without_live_proof_downgrades(self):
+        report = _reconcile_handoff().replace(
+            "- Linked issue live status: not verified in this session",
+            "- Linked issue live status: closed",
+        )
+        result = assess_final_report_validator(report, "reconcile_already_landed")
+        self.assertEqual(result["grade"], "downgraded")
+        self.assertTrue(
+            any(f["rule_id"] == "reconcile.linked_issue_live_status" for f in result["findings"])
+        )
+
+    def test_inventory_complete_requires_pagination_proof(self):
+        report = _reconcile_handoff() + "\nInventory complete for open PRs."
+        result = assess_final_report_validator(report, "reconcile_already_landed")
+        self.assertEqual(result["grade"], "downgraded")
+        self.assertTrue(
+            any(f["rule_id"] == "reconcile.inventory_pagination_proof" for f in result["findings"])
+        )
+
+
+class TestEntryPoint(unittest.TestCase):
+    def test_unknown_task_kind_blocks(self):
+        result = assess_final_report_validator("report", "unknown_mode")
+        self.assertTrue(result["blocked"])
+        self.assertEqual(result["findings"][0]["rule_id"], "shared.unknown_task_kind")
+
+    def test_review_proofs_reexport(self):
+        from review_proofs import assess_final_report_validator as exported
+
+        self.assertTrue(callable(exported))
+        result = exported(_review_handoff(), "review_pr")
+        self.assertIn("grade", result)
+
+    def test_supported_task_kinds_include_review_and_reconcile(self):
+        self.assertIn("review_pr", FINAL_REPORT_TASK_KINDS)
+        self.assertIn("reconcile_already_landed", FINAL_REPORT_TASK_KINDS)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_llm_workflow_split.py

@@ -118,6 +118,12 @@ def test_start_issue_template_references_work_issue_workflow():
     assert "workflows/work-issue.md" in text
 
 
+def test_final_report_validator_exported():
+    from review_proofs import assess_final_report_validator
+
+    assert callable(assess_final_report_validator)
+
+
 def test_create_issue_final_report_verifier_exported():
     from review_proofs import assess_create_issue_final_report