@ccoalm/ccl-skills 0.7.0 → 0.8.0

package/dist/assets/marketplace/plugins/ccl-skills/skills/skill-extraction-workflow/scripts/test_validate_extraction_review_state.sh

@@ -0,0 +1,1176 @@
+#!/usr/bin/env bash
+# Regression for receipt-bound extraction review terminal-state validation.
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+VALIDATOR="$SCRIPT_DIR/validate_extraction_review_state.py"
+[ -f "$VALIDATOR" ] || { echo "FAIL: validator not found: $VALIDATOR" >&2; exit 1; }
+
+TMP="$(mktemp -d "${TMPDIR:-/tmp}/extraction-review-state.XXXXXX")"
+trap 'rm -rf "$TMP"' EXIT
+
+python3 - "$VALIDATOR" "$TMP" <<'PY'
+import copy
+import hashlib
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+validator = Path(sys.argv[1])
+root = Path(sys.argv[2])
+A, B, C = (character * 40 for character in "abc")
+CANDIDATE = "d" * 64
+OTHER_CANDIDATE = "e" * 64
+CLOSED_DISPOSITIONS = {
+    "fixed",
+    "source_refuted",
+    "accepted_tradeoff",
+    "pre_existing_out_of_scope",
+}
+
+
+def canonical_hash(value):
+    encoded = json.dumps(
+        value, ensure_ascii=False, sort_keys=True, separators=(",", ":")
+    ).encode()
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def write_json(name, payload):
+    encoded = (json.dumps(payload, ensure_ascii=False, indent=2) + "\n").encode()
+    (root / name).write_bytes(encoded)
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def encode_duplicate_key(payload, field, first_value):
+    pairs = []
+    for key, value in payload.items():
+        if key == field:
+            pairs.append((key, first_value))
+        pairs.append((key, value))
+    return (
+        "{"
+        + ",".join(
+            f"{json.dumps(key)}:{json.dumps(value, ensure_ascii=False)}"
+            for key, value in pairs
+        )
+        + "}\n"
+    ).encode()
+
+
+def write_duplicate_json(name, payload, field, first_value):
+    encoded = encode_duplicate_key(payload, field, first_value)
+    (root / name).write_bytes(encoded)
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def write_base_evidence(name, base_sha, ref="refs/heads/dev"):
+    encoded = f"{base_sha}\t{ref}\n".encode()
+    (root / name).write_bytes(encoded)
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def occurrence_ref(occurrence):
+    return {
+        "receipt_sha256": occurrence["receipt_sha256"],
+        "finding_sha256": occurrence["finding_sha256"],
+    }
+
+
+def bind_disposition_evidence(name, occurrence, *, resolves=None, evidence=None):
+    if resolves is None:
+        resolves = [occurrence_ref(occurrence)]
+    if evidence is None:
+        evidence = [
+            f"synthetic {occurrence['disposition']} evidence for the current candidate"
+        ]
+    payload = {
+        "schema_version": 1,
+        "candidate_sha256": CANDIDATE,
+        "receipt_sha256": occurrence["receipt_sha256"],
+        "finding_sha256": occurrence["finding_sha256"],
+        "disposition": occurrence["disposition"],
+        "resolves_occurrences": copy.deepcopy(resolves),
+        "evidence": list(evidence),
+    }
+    evidence_hash = write_json(name, payload)
+    occurrence.update(
+        disposition_evidence_file=name,
+        disposition_evidence_sha256=evidence_hash,
+    )
+    return payload
+
+
+def clear_disposition_evidence(occurrence):
+    occurrence.pop("disposition_evidence_file", None)
+    occurrence.pop("disposition_evidence_sha256", None)
+
+
+def mutate_disposition_evidence(occurrence, mutator):
+    evidence_path = root / occurrence["disposition_evidence_file"]
+    payload = json.loads(evidence_path.read_text(encoding="utf-8"))
+    mutator(payload)
+    occurrence["disposition_evidence_sha256"] = write_json(
+        occurrence["disposition_evidence_file"], payload
+    )
+
+
+def duplicate_disposition_evidence_key(occurrence, field, first_value):
+    evidence_path = root / occurrence["disposition_evidence_file"]
+    payload = json.loads(evidence_path.read_text(encoding="utf-8"))
+    encoded = encode_duplicate_key(payload, field, first_value)
+    evidence_path.write_bytes(encoded)
+    occurrence["disposition_evidence_sha256"] = hashlib.sha256(encoded).hexdigest()
+
+
+def finding(number):
+    return {
+        "severity": "P1",
+        "title": f"receipt-bound finding {number}",
+        "body": f"failure path {number}",
+    }
+
+
+def make_fixture(
+    name,
+    *,
+    receipt_findings=None,
+    receipt_mutators=None,
+    completion=True,
+    completion_mutator=None,
+    budget=2,
+    base_shas=None,
+    closeout=None,
+    unmatched=0,
+    open_last=False,
+):
+    if receipt_findings is None:
+        receipt_findings = [[finding(1)], []]
+    receipt_mutators = receipt_mutators or {}
+    rounds = len(receipt_findings)
+    scope = {
+        "schema_version": 3,
+        "intent_sha256": "1" * 64,
+        "acceptance_sha256": "2" * 64,
+        "stage": "release",
+        "review_depth": "release",
+        "risk_tags": ["shared-gate"],
+        "challenge_budget": budget,
+        "wording_only_proof_sha256": None,
+        "wording_only_scope_sha256": None,
+    }
+    scope_sha = canonical_hash(scope)
+    receipts = []
+    receipt_hashes = []
+    receipt_refs = []
+    for index, findings in enumerate(receipt_findings, start=1):
+        mode = "review" if index == 1 else "challenge"
+        status = "findings" if findings else "passed"
+        remaining = 3 - index
+        if status == "findings":
+            state = "post_review_budget" if index == 3 else "findings_pending"
+        else:
+            state = "reviewed"
+        receipt = {
+            "schema_version": 3,
+            "mode": mode,
+            "status": status,
+            "findings": copy.deepcopy(findings),
+            "review_chain_tracked": True,
+            "review_chain_id": "extraction-chain",
+            "autonomous_review_index": index,
+            "autonomous_review_budget": 3,
+            "autonomous_reviews_remaining": remaining,
+            "autonomous_review_allowed": remaining > 0,
+            "challenge_index": 0 if index == 1 else index - 1,
+            "challenge_budget": budget,
+            "prior_review_result_sha256": list(receipt_hashes),
+            "candidate_sha256": CANDIDATE,
+            "packet_sha256": CANDIDATE,
+            "review_scope": copy.deepcopy(scope),
+            "review_scope_sha256": scope_sha,
+            "wording_only_proof_sha256": None,
+            "wording_only_scope": None,
+            "stage": "release",
+            "review_depth": "release",
+            "risk_tags": ["shared-gate"],
+            "review_state": state,
+            "human_decision_required": state == "post_review_budget",
+        }
+        mutator = receipt_mutators.get(index)
+        if mutator is not None:
+            mutator(receipt)
+        receipt_name = f"{name}-round-{index}.json"
+        receipt_hash = write_json(receipt_name, receipt)
+        receipts.append(receipt)
+        receipt_hashes.append(receipt_hash)
+        receipt_refs.append(
+            {"sequence": index, "file": receipt_name, "sha256": receipt_hash}
+        )
+
+    completion_ref = None
+    if completion:
+        final = receipts[-1]
+        complete = {
+            "schema_version": 3,
+            "mode": "complete",
+            "status": "passed",
+            "findings": [],
+            "review_chain_tracked": True,
+            "review_chain_id": final["review_chain_id"],
+            "autonomous_review_index": final["autonomous_review_index"],
+            "autonomous_review_budget": 3,
+            "autonomous_reviews_remaining": final["autonomous_reviews_remaining"],
+            "autonomous_review_allowed": False,
+            "challenge_budget": budget,
+            "prior_review_result_sha256": copy.deepcopy(
+                final["prior_review_result_sha256"]
+            ),
+            "candidate_sha256": final["candidate_sha256"],
+            "packet_sha256": final["packet_sha256"],
+            "review_scope": copy.deepcopy(final["review_scope"]),
+            "review_scope_sha256": final["review_scope_sha256"],
+            "wording_only_proof_sha256": None,
+            "wording_only_scope": None,
+            "stage": final["stage"],
+            "review_depth": final["review_depth"],
+            "risk_tags": copy.deepcopy(final["risk_tags"]),
+            "review_state": "self_reviewed",
+            "completion_gated": False,
+            "completion_review_result_sha256": receipt_hashes[-1],
+            "next_action": "complete",
+        }
+        if completion_mutator is not None:
+            completion_mutator(complete)
+        complete_name = f"{name}-complete.json"
+        complete_hash = write_json(complete_name, complete)
+        completion_ref = {"file": complete_name, "sha256": complete_hash}
+
+    if base_shas is None:
+        base_shas = [A] * rounds
+    attestations = []
+    for sequence, base_sha in enumerate(base_shas, start=1):
+        evidence_name = f"{name}-base-{sequence}.txt"
+        evidence_hash = write_base_evidence(evidence_name, base_sha)
+        attestations.append(
+            {
+                "sequence": sequence,
+                "remote": "origin",
+                "ref": "refs/heads/dev",
+                "sha": base_sha,
+                "confirmed_at": f"2026-08-29T12:{sequence:02d}:00Z",
+                "controller_receipt_sha256": (
+                    receipt_hashes[sequence - 1] if sequence <= rounds else None
+                ),
+                "evidence_file": evidence_name,
+                "evidence_sha256": evidence_hash,
+            }
+        )
+
+    occurrences = []
+    for receipt, receipt_hash in zip(receipts, receipt_hashes):
+        for item in receipt["findings"]:
+            occurrences.append(
+                {
+                    "receipt_sha256": receipt_hash,
+                    "finding_sha256": canonical_hash(item),
+                    "disposition": "fixed",
+                }
+            )
+    if open_last and occurrences:
+        occurrences[-1]["disposition"] = "open"
+    for occurrence_index, occurrence in enumerate(occurrences, start=1):
+        if occurrence["disposition"] in CLOSED_DISPOSITIONS:
+            bind_disposition_evidence(
+                f"{name}-disposition-{occurrence_index}.json",
+                occurrence,
+            )
+
+    classes = []
+    if occurrences:
+        sweep = None
+        if len(occurrences) >= 3:
+            searched_set = [
+                "skills/code-review",
+                "skills/skill-extraction-workflow",
+            ]
+            unmatched_items = [f"unmatched-{index}" for index in range(unmatched)]
+            manifest = {
+                "schema_version": 1,
+                "candidate_sha256": CANDIDATE,
+                "searched_set": searched_set,
+                "unmatched_instances": unmatched_items,
+            }
+            manifest_name = f"{name}-sweep.json"
+            manifest_hash = write_json(manifest_name, manifest)
+            sweep = {
+                "candidate_sha256": CANDIDATE,
+                "manifest_file": manifest_name,
+                "manifest_sha256": manifest_hash,
+                "searched_set": searched_set,
+                "unmatched_instances": unmatched,
+            }
+        classes.append(
+            {
+                "key": "receipt-binding",
+                "root_cause_predicate": "A closeout field is accepted without controller evidence.",
+                "affected_surface": "extraction closeout ledger",
+                "occurrences": occurrences,
+                "authoritative_sweep": sweep,
+            }
+        )
+
+    if closeout is None:
+        closeout = (
+            "ready_for_human_decision"
+            if completion
+            else "continuation_authorization_required"
+        )
+    controller_state = (
+        "self_reviewed" if completion else receipts[-1]["review_state"]
+    )
+    ledger = {
+        "schema_version": 3,
+        "candidate_sha256": CANDIDATE,
+        "controller_receipts": receipt_refs,
+        "completion_receipt": completion_ref,
+        "base_attestations": attestations,
+        "autonomous_round": rounds,
+        "controller_review_state": controller_state,
+        "finding_classes": classes,
+        "unreviewed_delta": [],
+        "closeout_state": closeout,
+    }
+    return {
+        "ledger": ledger,
+        "receipts": receipts,
+        "receipt_hashes": receipt_hashes,
+    }
+
+
+def invoke_encoded(name, encoded):
+    path = root / f"{name}-ledger.json"
+    path.write_bytes(encoded)
+    try:
+        return subprocess.run(
+            [sys.executable, str(validator), str(path)],
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+            text=True,
+            check=False,
+            timeout=5,
+        )
+    except subprocess.TimeoutExpired as exc:
+        raise AssertionError((name, "validator blocked", exc)) from exc
+
+
+def invoke(name, payload):
+    encoded = (json.dumps(payload, indent=2) + "\n").encode()
+    return invoke_encoded(name, encoded)
+
+
+def run(name, payload, expected_rc, token):
+    result = invoke(name, payload)
+    assert result.returncode == expected_rc, (name, result.returncode, result.stdout)
+    assert token in result.stdout, (name, token, result.stdout)
+    assert "Traceback" not in result.stdout, (name, result.stdout)
+
+
+ready = make_fixture("ready")
+run("ready", ready["ledger"], 0, "closeout_state=ready_for_human_decision")
+
+schema_v3 = make_fixture("schema-v3")
+schema_v3["ledger"]["schema_version"] = 3
+legacy_v2 = make_fixture("legacy-v2")
+legacy_v2["ledger"]["schema_version"] = 2
+schema_results = [
+    ("schema-v3", invoke("schema-v3", schema_v3["ledger"])),
+    ("legacy-v2", invoke("legacy-v2", legacy_v2["ledger"])),
+]
+assert [
+    (result.returncode, expected_token in result.stdout, "Traceback" in result.stdout)
+    for (name, result), expected_token in zip(
+        schema_results,
+        ("closeout_state=ready_for_human_decision", "schema_version must be 3"),
+        strict=True,
+    )
+] == [(0, True, False), (1, True, False)], [
+    (name, result.returncode, result.stdout) for name, result in schema_results
+]
+
+bad_digest = make_fixture("bad-digest")
+bad_digest["ledger"]["controller_receipts"][0]["sha256"] = "f" * 64
+run("bad-digest", bad_digest["ledger"], 1, "digest does not match")
+
+# JSON floats equal to the expected integers must not satisfy integer guards.
+float_schema = make_fixture("float-schema")
+float_schema["ledger"]["schema_version"] = 3.0
+run("float-schema", float_schema["ledger"], 1, "schema_version must be 3")
+
+float_budget = make_fixture(
+    "float-budget",
+    receipt_mutators={2: lambda row: row.update(challenge_budget=2.0)},
+)
+run("float-budget", float_budget["ledger"], 1, "challenge_budget must be 2")
+
+nan_schema = make_fixture("nan-schema")
+nan_schema["ledger"]["schema_version"] = float("nan")
+run("nan-schema", nan_schema["ledger"], 1, "non-standard JSON constant")
+
+control_chain = make_fixture(
+    "control-chain",
+    receipt_mutators={
+        2: lambda row: row.update(review_chain_id="chain\x1b[31mred")
+    },
+)
+run("control-chain", control_chain["ledger"], 1, "must not contain control characters")
+
+# C1 controls and Unicode line/paragraph separators forge diagnostic line
+# breaks just like C0 escapes; each must be rejected, not echoed.
+for label, hostile in (
+    ("c1-csi", "chain\x9bred"),
+    ("c1-nel", "chain\x85red"),
+    ("line-sep", "chain\u2028red"),
+    ("para-sep", "chain\u2029red"),
+):
+    hostile_chain = make_fixture(
+        f"hostile-{label}",
+        receipt_mutators={2: lambda row, value=hostile: row.update(review_chain_id=value)},
+    )
+    run(
+        f"hostile-{label}",
+        hostile_chain["ledger"],
+        1,
+        "must not contain control characters",
+    )
+
+# A float completion schema_version must not satisfy the completion guard.
+float_completion = make_fixture(
+    "float-completion",
+    completion_mutator=lambda row: row.update(schema_version=3.0),
+)
+run(
+    "float-completion",
+    float_completion["ledger"],
+    1,
+    "completion receipt must be a passed self_reviewed complete result",
+)
+
+wrong_chain = make_fixture(
+    "wrong-chain",
+    receipt_mutators={2: lambda row: row.update(review_chain_id="other-chain")},
+)
+run("wrong-chain", wrong_chain["ledger"], 1, "review_chain_id")
+
+wrong_mode = make_fixture(
+    "wrong-mode", receipt_mutators={2: lambda row: row.update(mode="review")}
+)
+run("wrong-mode", wrong_mode["ledger"], 1, "mode challenge")
+
+wrong_prior = make_fixture(
+    "wrong-prior",
+    receipt_mutators={
+        2: lambda row: row.update(prior_review_result_sha256=["f" * 64])
+    },
+)
+run("wrong-prior", wrong_prior["ledger"], 1, "prior_review_result_sha256")
+
+packet_mismatch = make_fixture(
+    "packet-mismatch",
+    receipt_mutators={2: lambda row: row.update(packet_sha256=OTHER_CANDIDATE)},
+)
+run("packet-mismatch", packet_mismatch["ledger"], 1, "packet_sha256 must equal candidate_sha256")
+
+def drift_scope(row):
+    row["review_scope"]["stage"] = "build"
+    row["review_scope_sha256"] = canonical_hash(row["review_scope"])
+    row["stage"] = "build"
+
+
+scope_mismatch = make_fixture("scope-mismatch", receipt_mutators={2: drift_scope})
+run("scope-mismatch", scope_mismatch["ledger"], 1, "review scope changed")
+
+budget_four = make_fixture("budget-four", budget=4)
+run("budget-four", budget_four["ledger"], 1, "challenge_budget must be 2")
+
+review_only = make_fixture("review-only", receipt_findings=[[]])
+run("review-only", review_only["ledger"], 1, "ready requires at least one tracked challenge")
+
+missing_complete = make_fixture("missing-complete")
+missing_complete["ledger"]["completion_receipt"] = None
+run("missing-complete", missing_complete["ledger"], 1, "requires a completion receipt")
+
+bad_complete_state = make_fixture(
+    "bad-complete-state",
+    completion_mutator=lambda row: row.update(review_state="reviewed"),
+)
+run("bad-complete-state", bad_complete_state["ledger"], 1, "passed self_reviewed")
+
+bad_complete_hash = make_fixture(
+    "bad-complete-hash",
+    completion_mutator=lambda row: row.update(
+        completion_review_result_sha256="f" * 64
+    ),
+)
+run("bad-complete-hash", bad_complete_hash["ledger"], 1, "completion_review_result_sha256")
+
+bad_complete_candidate = make_fixture(
+    "bad-complete-candidate",
+    completion_mutator=lambda row: row.update(
+        candidate_sha256=OTHER_CANDIDATE, packet_sha256=OTHER_CANDIDATE
+    ),
+)
+run("bad-complete-candidate", bad_complete_candidate["ledger"], 1, "exact final candidate")
+
+bad_complete_remaining = make_fixture(
+    "bad-complete-remaining",
+    completion_mutator=lambda row: row.update(autonomous_reviews_remaining=2),
+)
+run(
+    "bad-complete-remaining",
+    bad_complete_remaining["ledger"],
+    1,
+    "autonomous_reviews_remaining",
+)
+
+continuation = make_fixture(
+    "continuation",
+    receipt_findings=[[finding(1)], [], [finding(2)]],
+    completion=False,
+    open_last=True,
+)
+run(
+    "continuation",
+    continuation["ledger"],
+    0,
+    "closeout_state=continuation_authorization_required",
+)
+
+continuation_unconsumed_first_drift = make_fixture(
+    "continuation-unconsumed-first-drift",
+    receipt_findings=[[finding(1)], [], [finding(2)]],
+    completion=False,
+    open_last=True,
+    base_shas=[A, A, A, B],
+)
+run(
+    "continuation-unconsumed-first-drift",
+    continuation_unconsumed_first_drift["ledger"],
+    1,
+    "final controller receipt does not consume the latest attested base",
+)
+
+continuation_candidate = make_fixture(
+    "continuation-candidate",
+    receipt_findings=[[finding(1)], [], [finding(2)]],
+    completion=False,
+    open_last=True,
+)
+continuation_candidate["ledger"]["candidate_sha256"] = OTHER_CANDIDATE
+run(
+    "continuation-candidate",
+    continuation_candidate["ledger"],
+    1,
+    "controller receipt 1 does not bind the ledger candidate",
+)
+
+# A chain whose FIRST round reviewed a different candidate must not close out
+# a ledger for the final candidate, even when the final round binds it.
+drifted_round = make_fixture(
+    "drifted-round-candidate",
+    receipt_mutators={
+        1: lambda row: row.update(
+            candidate_sha256=OTHER_CANDIDATE, packet_sha256=OTHER_CANDIDATE
+        )
+    },
+)
+run(
+    "drifted-round-candidate",
+    drifted_round["ledger"],
+    1,
+    "controller receipt 1 does not bind the ledger candidate",
+)
+
+unknown_state = make_fixture(
+    "unknown-state",
+    receipt_findings=[[finding(1)], [], [finding(2)]],
+    receipt_mutators={3: lambda row: row.update(review_state="potato")},
+    completion=False,
+    open_last=True,
+)
+unknown_state["ledger"]["controller_review_state"] = "potato"
+run("unknown-state", unknown_state["ledger"], 1, "unknown controller review_state")
+
+passed_continuation = make_fixture(
+    "passed-continuation",
+    receipt_findings=[[finding(1)], [], []],
+    completion=False,
+)
+run("passed-continuation", passed_continuation["ledger"], 1, "round 3 findings in post_review_budget")
+
+bad_finding_hash = make_fixture("bad-finding-hash")
+bad_finding_hash["ledger"]["finding_classes"][0]["occurrences"][0][
+    "finding_sha256"
+] = "f" * 64
+run("bad-finding-hash", bad_finding_hash["ledger"], 1, "does not identify a finding")
+
+omitted_finding = make_fixture("omitted-finding")
+omitted_finding["ledger"]["finding_classes"] = []
+run("omitted-finding", omitted_finding["ledger"], 1, "omits controller findings")
+
+reordered_findings = make_fixture(
+    "reordered-findings", receipt_findings=[[finding(1), finding(2)], []]
+)
+reordered_findings["ledger"]["finding_classes"][0]["occurrences"].reverse()
+run("reordered-findings", reordered_findings["ledger"], 1, "not in controller receipt order")
+
+fixed_without_evidence = make_fixture("fixed-without-evidence")
+clear_disposition_evidence(
+    fixed_without_evidence["ledger"]["finding_classes"][0]["occurrences"][-1]
+)
+run(
+    "fixed-without-evidence",
+    fixed_without_evidence["ledger"],
+    1,
+    "disposition_evidence_file",
+)
+
+for disposition in ("accepted_tradeoff", "pre_existing_out_of_scope"):
+    case_name = f"{disposition.replace('_', '-')}-without-evidence"
+    without_evidence = make_fixture(case_name)
+    without_evidence_occurrence = without_evidence["ledger"]["finding_classes"][0][
+        "occurrences"
+    ][-1]
+    without_evidence_occurrence["disposition"] = disposition
+    clear_disposition_evidence(without_evidence_occurrence)
+    run(case_name, without_evidence["ledger"], 1, "disposition_evidence_file")
+
+    bound_case_name = f"{disposition.replace('_', '-')}-bound"
+    bound = make_fixture(bound_case_name)
+    bound_occurrence = bound["ledger"]["finding_classes"][0]["occurrences"][-1]
+    bound_occurrence["disposition"] = disposition
+    bind_disposition_evidence(
+        f"{bound_case_name}-disposition.json",
+        bound_occurrence,
+    )
+    run(bound_case_name, bound["ledger"], 0, "ready_for_human_decision")
+
+stale_disposition_evidence = make_fixture("stale-disposition-evidence")
+stale_disposition_evidence_occurrence = stale_disposition_evidence["ledger"][
+    "finding_classes"
+][0]["occurrences"][-1]
+mutate_disposition_evidence(
+    stale_disposition_evidence_occurrence,
+    lambda evidence: evidence.update(candidate_sha256=OTHER_CANDIDATE),
+)
+run(
+    "stale-disposition-evidence",
+    stale_disposition_evidence["ledger"],
+    1,
+    "stale for the current candidate",
+)
+
+wrong_disposition_evidence = make_fixture("wrong-disposition-evidence")
+wrong_disposition_evidence_occurrence = wrong_disposition_evidence["ledger"][
+    "finding_classes"
+][0]["occurrences"][-1]
+mutate_disposition_evidence(
+    wrong_disposition_evidence_occurrence,
+    lambda evidence: evidence.update(disposition="accepted_tradeoff"),
+)
+run(
+    "wrong-disposition-evidence",
+    wrong_disposition_evidence["ledger"],
+    1,
+    "does not bind its disposition",
+)
+
+duplicate_key_cases = []
+for case_name, field, first_value in (
+    ("duplicate-candidate-evidence", "candidate_sha256", OTHER_CANDIDATE),
+    ("duplicate-disposition-evidence", "disposition", "accepted_tradeoff"),
+):
+    duplicate_case = make_fixture(case_name)
+    duplicate_occurrence = duplicate_case["ledger"]["finding_classes"][0][
+        "occurrences"
+    ][-1]
+    duplicate_disposition_evidence_key(duplicate_occurrence, field, first_value)
+    duplicate_result = invoke(case_name, duplicate_case["ledger"])
+    duplicate_key_cases.append(
+        (
+            case_name,
+            duplicate_result.returncode,
+            f"duplicate object key: {field}" in duplicate_result.stdout,
+            "Traceback" in duplicate_result.stdout,
+            duplicate_result.stdout,
+        )
+    )
+assert [item[1:4] for item in duplicate_key_cases] == [
+    (1, True, False),
+    (1, True, False),
+], duplicate_key_cases
+
+wrong_receipt_evidence = make_fixture("wrong-receipt-evidence")
+wrong_receipt_evidence_occurrence = wrong_receipt_evidence["ledger"][
+    "finding_classes"
+][0]["occurrences"][-1]
+mutate_disposition_evidence(
+    wrong_receipt_evidence_occurrence,
+    lambda evidence: evidence.update(receipt_sha256="f" * 64),
+)
+run(
+    "wrong-receipt-evidence",
+    wrong_receipt_evidence["ledger"],
+    1,
+    "does not bind its controller receipt",
+)
+
+wrong_finding_evidence = make_fixture("wrong-finding-evidence")
+wrong_finding_evidence_occurrence = wrong_finding_evidence["ledger"][
+    "finding_classes"
+][0]["occurrences"][-1]
+mutate_disposition_evidence(
+    wrong_finding_evidence_occurrence,
+    lambda evidence: evidence.update(finding_sha256="f" * 64),
+)
+run(
+    "wrong-finding-evidence",
+    wrong_finding_evidence["ledger"],
+    1,
+    "does not bind its controller finding",
+)
+
+historical_open = make_fixture(
+    "historical-open",
+    receipt_findings=[[finding(1)], [finding(2)], []],
+)
+historical_open["ledger"]["finding_classes"][0]["occurrences"][0][
+    "disposition"
+] = "open"
+clear_disposition_evidence(
+    historical_open["ledger"]["finding_classes"][0]["occurrences"][0]
+)
+run("historical-open", historical_open["ledger"], 1, "unresolved finding occurrence")
+
+historical_open_linked = make_fixture(
+    "historical-open-linked",
+    receipt_findings=[[finding(1)], [finding(2)], []],
+)
+historical_open_linked_occurrences = historical_open_linked["ledger"][
+    "finding_classes"
+][0]["occurrences"]
+historical_open_linked_occurrences[0]["disposition"] = "open"
+clear_disposition_evidence(historical_open_linked_occurrences[0])
+bind_disposition_evidence(
+    "historical-open-linked-disposition.json",
+    historical_open_linked_occurrences[1],
+    resolves=[
+        occurrence_ref(historical_open_linked_occurrences[0]),
+        occurrence_ref(historical_open_linked_occurrences[1]),
+    ],
+)
+run(
+    "historical-open-linked",
+    historical_open_linked["ledger"],
+    0,
+    "ready_for_human_decision",
+)
+
+source_refuted_without_evidence = make_fixture("source-refuted-without-evidence")
+source_refuted_without_evidence_occurrence = source_refuted_without_evidence["ledger"][
+    "finding_classes"
+][0]["occurrences"][-1]
+source_refuted_without_evidence_occurrence["disposition"] = "source_refuted"
+clear_disposition_evidence(source_refuted_without_evidence_occurrence)
+run(
+    "source-refuted-without-evidence",
+    source_refuted_without_evidence["ledger"],
+    1,
+    "disposition_evidence_file",
+)
+
+source_refuted = make_fixture("source-refuted")
+source_refuted_occurrence = source_refuted["ledger"]["finding_classes"][0][
+    "occurrences"
+][-1]
+source_refuted_occurrence["disposition"] = "source_refuted"
+bind_disposition_evidence(
+    "source-refuted-evidence.json",
+    source_refuted_occurrence,
+    evidence=["synthetic first-hand source evidence"],
+)
+run("source-refuted", source_refuted["ledger"], 0, "ready_for_human_decision")
+
+source_refuted_bad_hash = copy.deepcopy(source_refuted["ledger"])
+source_refuted_bad_hash["finding_classes"][0]["occurrences"][-1][
+    "disposition_evidence_sha256"
+] = "f" * 64
+run(
+    "source-refuted-bad-hash",
+    source_refuted_bad_hash,
+    1,
+    "digest does not match",
+)
+
+needs_human = make_fixture("needs-human")
+needs_human["ledger"]["finding_classes"][0]["occurrences"][-1][
+    "disposition"
+] = "needs_human_decision"
+clear_disposition_evidence(
+    needs_human["ledger"]["finding_classes"][0]["occurrences"][-1]
+)
+run("needs-human", needs_human["ledger"], 1, "unresolved finding occurrence")
+
+third_occurrence = make_fixture(
+    "third-occurrence",
+    receipt_findings=[[finding(1)], [finding(2)], [finding(3)]],
+    completion=False,
+    unmatched=1,
+    open_last=True,
+)
+run("third-occurrence", third_occurrence["ledger"], 0, "continuation_authorization_required")
+
+split_predicate = copy.deepcopy(third_occurrence["ledger"])
+original_class = split_predicate["finding_classes"][0]
+predicate_variants = [
+    "A closeout field is accepted without controller evidence.",
+    "a   closeout field is accepted without controller evidence.",
+    "A closeout field is accepted without controller evidence.",
+]
+split_predicate["finding_classes"] = [
+    {
+        "key": f"receipt-binding-{index}",
+        "root_cause_predicate": predicate_variants[index - 1],
+        "affected_surface": original_class["affected_surface"],
+        "occurrences": [copy.deepcopy(occurrence)],
+        "authoritative_sweep": None,
+    }
+    for index, occurrence in enumerate(original_class["occurrences"], start=1)
+]
+run(
+    "split-predicate",
+    split_predicate,
+    1,
+    "duplicate normalized root_cause_predicate",
+)
+
+distinct_predicates = copy.deepcopy(split_predicate)
+for index, row in enumerate(distinct_predicates["finding_classes"], start=1):
+    row["root_cause_predicate"] = (
+        f"Distinct controller-evidence failure predicate {index}."
+    )
+run(
+    "distinct-predicates",
+    distinct_predicates,
+    0,
+    "closeout_state=continuation_authorization_required",
+)
+
+missing_sweep = copy.deepcopy(third_occurrence["ledger"])
+missing_sweep["finding_classes"][0]["authoritative_sweep"] = None
+run("missing-sweep", missing_sweep, 1, "authoritative_sweep")
+
+bad_sweep_set = copy.deepcopy(third_occurrence["ledger"])
+bad_sweep_set["finding_classes"][0]["authoritative_sweep"]["searched_set"] = [
+    "skills/code-review"
+]
+run("bad-sweep-set", bad_sweep_set, 1, "searched_set does not match")
+
+ready_unmatched = make_fixture(
+    "ready-unmatched",
+    receipt_findings=[[finding(1), finding(2), finding(3)], []],
+    unmatched=1,
+)
+run("ready-unmatched", ready_unmatched["ledger"], 1, "ready sweep must report zero unmatched")
+
+wrong_remote = make_fixture("wrong-remote")
+wrong_remote["ledger"]["base_attestations"][1]["remote"] = "upstream"
+run("wrong-remote", wrong_remote["ledger"], 1, "same remote and ref")
+
+bad_sequence = make_fixture("bad-sequence")
+bad_sequence["ledger"]["base_attestations"][1]["sequence"] = 3
+run("bad-sequence", bad_sequence["ledger"], 1, "sequence must be contiguous")
+
+bad_time = make_fixture("bad-time")
+bad_time["ledger"]["base_attestations"][1]["confirmed_at"] = "2026-08-29T12:01:00Z"
+run("bad-time", bad_time["ledger"], 1, "strictly increase")
+
+invalid_time = make_fixture("invalid-time")
+invalid_time["ledger"]["base_attestations"][0]["confirmed_at"] = "2026-08-29 12:00:00"
+run("invalid-time", invalid_time["ledger"], 1, "strict RFC3339")
+
+missing_mapping = make_fixture("missing-mapping")
+missing_mapping["ledger"]["base_attestations"][1]["controller_receipt_sha256"] = None
+run("missing-mapping", missing_mapping["ledger"], 1, "map every controller receipt")
+
+consumed_first_drift = make_fixture("consumed-first-drift", base_shas=[A, B])
+run(
+    "consumed-first-drift",
+    consumed_first_drift["ledger"],
+    0,
+    "closeout_state=ready_for_human_decision",
+)
+
+stable_trailing_recheck = make_fixture(
+    "stable-trailing-recheck", base_shas=[A, A, A]
+)
+run(
+    "stable-trailing-recheck",
+    stable_trailing_recheck["ledger"],
+    0,
+    "closeout_state=ready_for_human_decision",
+)
+
+unconsumed_first_drift = make_fixture(
+    "unconsumed-first-drift", base_shas=[A, A, B]
+)
+run(
+    "unconsumed-first-drift",
+    unconsumed_first_drift["ledger"],
+    1,
+    "final controller receipt does not consume the latest attested base",
+)
+
+bad_evidence = make_fixture("bad-evidence")
+bad_evidence["ledger"]["base_attestations"][0]["evidence_sha256"] = "f" * 64
+run("bad-evidence", bad_evidence["ledger"], 1, "digest does not match")
+
+bad_evidence_content = make_fixture("bad-evidence-content")
+bad_evidence_row = bad_evidence_content["ledger"]["base_attestations"][0]
+bad_evidence_path = root / bad_evidence_row["evidence_file"]
+bad_evidence_path.write_text(f"{A}\trefs/heads/other\n", encoding="utf-8")
+bad_evidence_row["evidence_sha256"] = hashlib.sha256(
+    bad_evidence_path.read_bytes()
+).hexdigest()
+run(
+    "bad-evidence-content",
+    bad_evidence_content["ledger"],
+    1,
+    "canonical ls-remote output",
+)
+
+surrogate = make_fixture("surrogate")
+surrogate["ledger"]["controller_review_state"] = "\ud800"
+run("surrogate", surrogate["ledger"], 1, "valid UTF-8 text")
+
+fifo = make_fixture("fifo")
+fifo_name = "controller-receipt.fifo"
+os.mkfifo(root / fifo_name)
+fifo["ledger"]["controller_receipts"][0]["file"] = fifo_name
+fifo["ledger"]["controller_receipts"][0]["sha256"] = "f" * 64
+run("fifo", fifo["ledger"], 1, "singly linked regular file")
+
+second_drift_wrong = make_fixture("second-drift-wrong", base_shas=[A, B, A])
+run("second-drift-wrong", second_drift_wrong["ledger"], 1, "second base drift")
+
+flip_flop = make_fixture(
+    "flip-flop",
+    completion=False,
+    base_shas=[A, B, A],
+    closeout="baseline_race",
+)
+flip_flop["ledger"]["unreviewed_delta"] = ["target returned to its first SHA"]
+run("flip-flop", flip_flop["ledger"], 0, "base_changes=2")
+
+one_round_race = make_fixture(
+    "one-round-race",
+    receipt_findings=[[]],
+    completion=False,
+    base_shas=[A, B, C],
+    closeout="baseline_race",
+)
+one_round_race["ledger"]["unreviewed_delta"] = ["second target drift after round 1"]
+run("one-round-race", one_round_race["ledger"], 0, "base_changes=2")
+
+post_race_round = make_fixture(
+    "post-race-round",
+    completion=False,
+    base_shas=[A, B, C],
+    closeout="baseline_race",
+)
+post_race_round["ledger"]["base_attestations"][1][
+    "controller_receipt_sha256"
+] = None
+post_race_round["ledger"]["base_attestations"][2][
+    "controller_receipt_sha256"
+] = post_race_round["receipt_hashes"][1]
+post_race_round["ledger"]["unreviewed_delta"] = [
+    "round 2 was started after the second target drift"
+]
+run(
+    "post-race-round",
+    post_race_round["ledger"],
+    1,
+    "controller receipt is mapped at or after the second base drift",
+)
+
+race_with_complete = make_fixture(
+    "race-with-complete", base_shas=[A, B, C], closeout="baseline_race"
+)
+race_with_complete["ledger"]["unreviewed_delta"] = ["second target drift"]
+run("race-with-complete", race_with_complete["ledger"], 1, "cannot carry a completion receipt")
+
+race_without_delta = make_fixture(
+    "race-without-delta",
+    completion=False,
+    base_shas=[A, B, C],
+    closeout="baseline_race",
+)
+run("race-without-delta", race_without_delta["ledger"], 1, "non-empty unreviewed_delta")
+
+race_candidate = make_fixture(
+    "race-candidate",
+    completion=False,
+    base_shas=[A, B, C],
+    closeout="baseline_race",
+)
+race_candidate["ledger"]["candidate_sha256"] = OTHER_CANDIDATE
+race_candidate["ledger"]["unreviewed_delta"] = ["second target drift"]
+run(
+    "race-candidate",
+    race_candidate["ledger"],
+    1,
+    "controller receipt 1 does not bind the ledger candidate",
+)
+
+false_race = make_fixture(
+    "false-race", completion=False, closeout="baseline_race"
+)
+run("false-race", false_race["ledger"], 1, "baseline_race requires two ordered base changes")
+
+needs_human_resolution_results = []
+for closing_disposition in sorted(CLOSED_DISPOSITIONS):
+    case_name = f"historical-needs-human-{closing_disposition.replace('_', '-')}"
+    historical_needs_human_linked = make_fixture(
+        case_name,
+        receipt_findings=[[finding(1)], [finding(2)], []],
+    )
+    historical_needs_human_occurrences = historical_needs_human_linked["ledger"][
+        "finding_classes"
+    ][0]["occurrences"]
+    historical_needs_human_occurrences[0]["disposition"] = "needs_human_decision"
+    clear_disposition_evidence(historical_needs_human_occurrences[0])
+    historical_needs_human_occurrences[1]["disposition"] = closing_disposition
+    bind_disposition_evidence(
+        f"{case_name}-disposition.json",
+        historical_needs_human_occurrences[1],
+        resolves=[
+            occurrence_ref(historical_needs_human_occurrences[0]),
+            occurrence_ref(historical_needs_human_occurrences[1]),
+        ],
+    )
+    needs_human_resolution_results.append(
+        (case_name, invoke(case_name, historical_needs_human_linked["ledger"]))
+    )
+
+duplicate_ledger = make_fixture("duplicate-ledger")
+duplicate_ledger_result = invoke_encoded(
+    "duplicate-ledger",
+    encode_duplicate_key(duplicate_ledger["ledger"], "schema_version", 2),
+)
+
+duplicate_controller = make_fixture("duplicate-controller")
+duplicate_controller_ledger = duplicate_controller["ledger"]
+duplicate_controller_name = duplicate_controller_ledger["controller_receipts"][-1][
+    "file"
+]
+duplicate_controller_payload = json.loads(
+    (root / duplicate_controller_name).read_text(encoding="utf-8")
+)
+duplicate_controller_hash = write_duplicate_json(
+    duplicate_controller_name,
+    duplicate_controller_payload,
+    "status",
+    "findings",
+)
+duplicate_controller_ledger["controller_receipts"][-1][
+    "sha256"
+] = duplicate_controller_hash
+duplicate_controller_ledger["base_attestations"][-1][
+    "controller_receipt_sha256"
+] = duplicate_controller_hash
+duplicate_controller_complete_name = duplicate_controller_ledger[
+    "completion_receipt"
+]["file"]
+duplicate_controller_complete = json.loads(
+    (root / duplicate_controller_complete_name).read_text(encoding="utf-8")
+)
+duplicate_controller_complete[
+    "completion_review_result_sha256"
+] = duplicate_controller_hash
+duplicate_controller_ledger["completion_receipt"]["sha256"] = write_json(
+    duplicate_controller_complete_name,
+    duplicate_controller_complete,
+)
+duplicate_controller_result = invoke(
+    "duplicate-controller",
+    duplicate_controller_ledger,
+)
+
+duplicate_completion = make_fixture("duplicate-completion")
+duplicate_completion_ledger = duplicate_completion["ledger"]
+duplicate_completion_name = duplicate_completion_ledger["completion_receipt"]["file"]
+duplicate_completion_payload = json.loads(
+    (root / duplicate_completion_name).read_text(encoding="utf-8")
+)
+duplicate_completion_ledger["completion_receipt"]["sha256"] = write_duplicate_json(
+    duplicate_completion_name,
+    duplicate_completion_payload,
+    "review_state",
+    "reviewed",
+)
+duplicate_completion_result = invoke(
+    "duplicate-completion",
+    duplicate_completion_ledger,
+)
+
+duplicate_sweep = make_fixture(
+    "duplicate-sweep",
+    receipt_findings=[[finding(1)], [finding(2)], [finding(3)]],
+    completion=False,
+    open_last=True,
+)
+duplicate_sweep_ledger = duplicate_sweep["ledger"]
+duplicate_sweep_row = duplicate_sweep_ledger["finding_classes"][0][
+    "authoritative_sweep"
+]
+duplicate_sweep_payload = json.loads(
+    (root / duplicate_sweep_row["manifest_file"]).read_text(encoding="utf-8")
+)
+duplicate_sweep_row["manifest_sha256"] = write_duplicate_json(
+    duplicate_sweep_row["manifest_file"],
+    duplicate_sweep_payload,
+    "candidate_sha256",
+    OTHER_CANDIDATE,
+)
+duplicate_sweep_result = invoke("duplicate-sweep", duplicate_sweep_ledger)
+
+new_regressions = [
+    *[
+        (
+            case_name,
+            result,
+            "cannot resolve a needs_human_decision occurrence",
+        )
+        for case_name, result in needs_human_resolution_results
+    ],
+    ("duplicate-ledger", duplicate_ledger_result, "duplicate object key: schema_version"),
+    ("duplicate-controller", duplicate_controller_result, "duplicate object key: status"),
+    (
+        "duplicate-completion",
+        duplicate_completion_result,
+        "duplicate object key: review_state",
+    ),
+    ("duplicate-sweep", duplicate_sweep_result, "duplicate object key: candidate_sha256"),
+]
+assert [
+    (name, result.returncode, token in result.stdout, "Traceback" in result.stdout)
+    for name, result, token in new_regressions
+] == [(name, 1, True, False) for name, _result, _token in new_regressions], [
+    (name, result.returncode, result.stdout) for name, result, _token in new_regressions
+]
+
+print("test_validate_extraction_review_state: ok")
+PY