Add value-safe verification and audit reporting

Assistant: codex
Assistant-Model: gpt-5.6-sol
Assistant-Session: 01a0217e-8c4c-7383-be6b-f50a6e485306

ProductRequirementsDocument.md

@@ -201,7 +201,7 @@ secrets-engine plan <decision-or-ccr-id>
 secrets-engine apply <decision-or-ccr-id> --stage <build|test|prod>
 secrets-engine provision <catalog-id> --stage <stage> --from-file <path>
 secrets-engine verify <catalog-id> --positive
-secrets-engine verify <catalog-id> --negative
+secrets-engine verify <catalog-id> --negative --negative-token-file <path>
 secrets-engine exec --catalog <catalog-id> -- <command...>
 secrets-engine revoke <catalog-id-or-lease>
 ```

README.md

@@ -7,7 +7,8 @@ and production stages.
 OpenBao remains the custody and enforcement backend. `secrets-engine` owns the
 operator and agent interaction model: catalog, decision checks, plan/apply,
 guarded provisioning, verification, delivery, evidence, lifecycle metadata, and
-native-access deactivation.
+native-access deactivation. Local evidence can be inspected through an
+allowlisted per-lane `audit` summary without exposing record detail.
 
 ## Start Here
 

SCOPE.md

@@ -58,14 +58,17 @@ provider, credential broker, or general secrets API.
 - KV positive verification logs in through the lane AppRole and reports field
   presence without printing values. With no `--field`, it checks every declared
   field; an explicit `--field` narrows the check.
-- KV negative verification checks that a fixed invalid token cannot read the
+- KV negative verification requires a real unrelated token supplied through a
-  path.
+  mode-0600 file outside Git and checks that it cannot read the path. Missing
+  unrelated identity material fails closed without a backend probe.
 - Auth-capability verification checks the AppRole token's capabilities on exact
   allowlisted and denial-probe paths.
 
-These are bounded policy/presence probes. They do not yet prove denial for a
+These are bounded policy/presence probes. A throwaway OpenBao integration test
-real unrelated workload identity, exercise a provider operation, or produce
+proves an overlapping unrelated policy is detected. Production still needs a
-OpenBao audit-request correlation.
+reviewed owner/source for each unrelated identity; the engine does not mint or
+select that identity itself. Verification does not yet exercise a provider
+operation or produce OpenBao audit-request correlation.
 
 ### Exec-time delivery
 
@@ -104,12 +107,19 @@ cannot be recovered through that AppRole.
   status, metadata/value-presence booleans, missing declared field names,
   readiness, and a safe next command. Every declared field must be present.
 - Records scrubbed local JSONL evidence and posts a minimal State Hub progress
-  event on a best-effort basis.
+  event on a best-effort basis. Each requested State Hub delivery receives an
+  append-only local `delivered`, `failed`, or `skipped-no-topic` companion
+  record.
+- `audit` summarizes allowlisted local lane evidence: action/result counts,
+  canonical decision references, session cleanup outcomes, and State Hub
+  delivery outcomes. It never re-emits arbitrary evidence detail.
 - Keeps OpenBao audit logs as the backend source of truth.
 
-State Hub evidence delivery is not durable or transactional, and route does not
+State Hub evidence delivery is not queued or transactional; the local receipt
-replace authorization or unrelated-identity denial evidence. Treat route output
+makes failure visible but does not replay it. Route and audit do not replace
-as operational guidance, not complete attestation for high-risk lanes.
+authorization, OpenBao audit logs, or unrelated-identity denial evidence.
+Treat their output as operational guidance, not complete attestation for
+high-risk lanes.
 
 ### Revocation currently available
 
@@ -145,6 +155,7 @@ secrets-engine policy publication
 secrets-engine route
 secrets-engine revoke [--dry-run]
 secrets-engine lifecycle suspend|deactivate|destroy
+secrets-engine audit <catalog-id> [--json]
 ```
 
 The implemented exec adapters are `exec-env` and `npm-config`. `read-check` is
@@ -254,7 +265,9 @@ keywords: [openbao, approle, auth-capability, handoff, least-privilege]
 type: governance
 title: Non-secret routing and evidence pointers
 description: Reports decision/readiness metadata and records scrubbed local and best-effort State Hub
-  evidence without returning secret values. Readiness covers every declared field, but this is not a
+  evidence without returning secret values. Append-only delivery receipts expose State Hub failures,
-  durable audit store or a real-unrelated-identity denial attestation.
+  and an allowlisted audit command summarizes lane operations and cleanup. KV verification can attest a
+  supplied real unrelated identity, but the engine does not own identity selection. This is not a durable
+  remote audit store.
 keywords: [routing, evidence, state-hub, audit, secrets]
 ```

docs/cli.md

@@ -48,7 +48,7 @@ secrets-engine decision inspect <decision-or-ccr-id>
 secrets-engine plan  <ref> --stage <build|test|prod>
 secrets-engine apply <ref> --stage <stage> [--dry-run] [--bootstrap-token-file F]
 secrets-engine provision <catalog-id> --stage <stage> --field NAME (--from-file F | --generate)
-secrets-engine verify <catalog-id> [--field NAME] [--positive] [--negative]
+secrets-engine verify <catalog-id> [--field NAME] [--positive] [--negative] [--negative-token-file F]
 secrets-engine handoff <catalog-id> --stage <stage> --role-id-file F --secret-id-file F
 secrets-engine exec --catalog <catalog-id> [--field NAME] [--mode auto|npm-config|exec-env] -- CMD...
 secrets-engine policy publication <catalog-id>
@@ -57,6 +57,7 @@ secrets-engine revoke <catalog-id> [--dry-run]
 secrets-engine lifecycle suspend <catalog-id> [--dry-run]
 secrets-engine lifecycle deactivate <catalog-id> [--dry-run]
 secrets-engine lifecycle destroy <catalog-id> [--dry-run] [--confirm-destroy <catalog-id>]
+secrets-engine audit <catalog-id> [--json]
 ```
 
 `policy publication` resolves a lane's effective publication scope and the env
@@ -81,7 +82,9 @@ CAS-aware create/patch behavior with a strict temporary input reference: values
 are absent from argv, sibling fields are preserved, and stale writes fail.
 
 With no `--field`, `verify` checks every declared KV field positively and runs
-one path-level negative probe. `route` likewise requires every declared field
+one path-level negative probe. KV denial requires `--negative-token-file` with a
+real unrelated identity's mode-0600 token file outside Git; absence fails the
+check without calling OpenBao. `route` likewise requires every declared field
 and reports only missing field names. An explicit `--field` narrows positive
 verification; production use will bind such subsets to the action approval.
 
@@ -102,6 +105,13 @@ Exec and verification AppRole logins are scoped sessions. The issued token
 self-revokes on every exit path before exec starts (or when verification ends),
 and evidence stores only a short accessor fingerprint plus cleanup outcome.
 
+`audit` is a read-only, local evidence summary. It reports action and result
+counts, canonical decision references, session cleanup outcomes, and State Hub
+delivery outcomes for one cataloged lane. Its parser allowlists those fields and
+does not echo arbitrary JSONL detail. State Hub failures are recorded locally as
+append-only companion receipts; they are visible but are not automatically
+replayed.
+
 ## Exit codes
 
 | Code | Meaning |
@@ -177,7 +187,8 @@ secrets-engine apply whynot-design-npm-publish --stage prod
 secrets-engine provision whynot-design-npm-publish --stage prod \
   --field npm_token --from-file ~/.secrets-engine/whynot.token
 # 5. prove access without printing the value
-secrets-engine verify whynot-design-npm-publish --field npm_token --positive --negative
+secrets-engine verify whynot-design-npm-publish --field npm_token --positive --negative \
+  --negative-token-file /secure/path/unrelated.token
 # 6. publish with the token injected into the child only
 secrets-engine exec --catalog whynot-design-npm-publish -- npm publish
 ```

docs/hardening-backlog.md

@@ -57,8 +57,10 @@ contents in this repo.
 
 ## H5 — Audit report command
 
-- `secrets-engine audit <catalog-id>` summarizing non-secret evidence (decision,
+- Implemented: `secrets-engine audit <catalog-id>` summarizes allowlisted local
-  applies, provisions, verifications, execs, revokes) for a lane.
+  action/result, decision, session-cleanup, and State Hub delivery evidence.
+- Remaining: durable/queued State Hub delivery and OpenBao audit request-id
+  correlation.
 
 ## H6 — API service mode
 

docs/whynot-design-real-publish-closeout.md

@@ -70,9 +70,14 @@ BAO_ADDR=https://bao.coulomb.social \
 BAO_ADDR=https://bao.coulomb.social \
   secrets-engine verify whynot-design-npm-publish \
   --field npm_token --positive --negative \
+  --negative-token-file /secure/path/reviewed-unrelated.token \
   --bootstrap-token-file ~/.secrets-engine/bootstrap/prod-whynot-design.token
 ```
 
+The unrelated token must represent a reviewed identity outside the lane policy;
+it is not the bootstrap token and must be held in a separate mode-0600 file
+outside Git.
+
 6. In the `whynot-design` repo, coordinate a real version bump. Run the publish
    through child-only injection:
 

scripts/demo-e2e.sh

@@ -21,6 +21,7 @@ export BAO_ADDR="http://127.0.0.1:8270"
 export BAO_TOKEN="se-demo-root"
 # Keep secret material OUTSIDE the repo worktree (engine enforces this).
 TOKENFILE="$WORK/whynot.token"
+NEGATIVE_TOKENFILE="$WORK/unrelated.token"
 
 cleanup() {
   set +e
@@ -59,7 +60,15 @@ secrets-engine provision whynot-design-npm-publish --stage prod \
 
 echo
 echo "### 4. verify positive (approved consumer can read) + negative (others denied)"
-secrets-engine verify whynot-design-npm-publish --field npm_token --positive --negative | sed 's/^/    /'
+"$BAO_BIN" policy write se-demo-unrelated - <<'EOF' >/dev/null
+path "secret/data/prod/coulomb/whynot-design/npm" {
+  capabilities = ["deny"]
+}
+EOF
+umask 077
+"$BAO_BIN" token create -field=token -policy=se-demo-unrelated > "$NEGATIVE_TOKENFILE"
+secrets-engine verify whynot-design-npm-publish --field npm_token \
+  --positive --negative --negative-token-file "$NEGATIVE_TOKENFILE" | sed 's/^/    /'
 
 echo
 echo "### 5. exec-time delivery: child sees the token via a temp npmrc; parent never does"

src/secrets_engine/audit.py

@@ -0,0 +1,167 @@
+"""Allowlisted, non-secret summaries over local append-only evidence."""
+from __future__ import annotations
+
+import json
+import re
+import uuid
+from collections import Counter
+from dataclasses import asdict, dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+_LABEL = re.compile(r"^[a-z][a-z0-9-]{0,79}$")
+_DECISION_REF = re.compile(r"^[A-Z][A-Z0-9-]{2,80}$")
+_DELIVERY_RESULTS = {"delivered", "failed", "skipped-no-topic"}
+_VERIFY_RESULT = re.compile(r"^(positive|negative):(pass|fail)$")
+_ERROR_RESULT = re.compile(
+    r"^failed-(Catalog|Decision|PolicyGuard|Backend|Provisioning|Verification|Delivery)Error$"
+)
+
+
+def _safe_label(value: object) -> str:
+    text = value if isinstance(value, str) else ""
+    return text if _LABEL.fullmatch(text) else "invalid-label"
+
+
+def _safe_result(value: object) -> str:
+    text = value if isinstance(value, str) else ""
+    if (
+        _LABEL.fullmatch(text)
+        or _VERIFY_RESULT.fullmatch(text)
+        or _ERROR_RESULT.fullmatch(text)
+    ):
+        return text
+    return "invalid-label"
+
+
+def _safe_decision_ref(value: object) -> str:
+    text = value if isinstance(value, str) else ""
+    if not text:
+        return ""
+    try:
+        return str(uuid.UUID(text))
+    except ValueError:
+        return text if _DECISION_REF.fullmatch(text) else ""
+
+
+def _safe_timestamp(value: object) -> str:
+    if not isinstance(value, str):
+        return ""
+    try:
+        parsed = datetime.fromisoformat(value.replace("Z", "+00:00"))
+        if parsed.tzinfo is None:
+            return ""
+        return parsed.astimezone(timezone.utc).isoformat()
+    except ValueError:
+        return ""
+
+
+@dataclass(frozen=True)
+class LaneAuditSummary:
+    catalog_id: str
+    operation_records: int
+    malformed_records: int
+    first_ts: str
+    last_ts: str
+    actions: dict[str, int]
+    results: dict[str, int]
+    decision_refs: list[str]
+    session_cleanup: dict[str, int]
+    hub_delivery: dict[str, int]
+
+    def to_json(self) -> dict[str, Any]:
+        return asdict(self)
+
+    def render(self) -> str:
+        lines = [
+            f"Lane audit summary for '{self.catalog_id}'",
+            f"  operation records: {self.operation_records}",
+            f"  malformed records: {self.malformed_records}",
+            f"  first: {self.first_ts or 'n/a'}",
+            f"  last:  {self.last_ts or 'n/a'}",
+            "  actions: " + _render_counts(self.actions),
+            "  results: " + _render_counts(self.results),
+            "  decisions: " + (", ".join(self.decision_refs) or "none"),
+            "  session cleanup: " + _render_counts(self.session_cleanup),
+            "  hub delivery: " + _render_counts(self.hub_delivery),
+        ]
+        return "\n".join(lines)
+
+
+def _render_counts(counts: dict[str, int]) -> str:
+    return ", ".join(f"{key}={value}" for key, value in sorted(counts.items())) or "none"
+
+
+def summarize_lane_evidence(evidence_dir: Path, catalog_id: str) -> LaneAuditSummary:
+    """Summarize one lane without returning arbitrary record fields or detail."""
+    actions: Counter[str] = Counter()
+    results: Counter[str] = Counter()
+    cleanup: Counter[str] = Counter()
+    hub_delivery: Counter[str] = Counter()
+    decisions: set[str] = set()
+    timestamps: list[str] = []
+    malformed = 0
+    operation_records = 0
+
+    for path in sorted(Path(evidence_dir).glob("evidence-*.jsonl")):
+        try:
+            lines = path.read_text(encoding="utf-8").splitlines()
+        except OSError:
+            malformed += 1
+            continue
+        for line in lines:
+            try:
+                record = json.loads(line)
+            except (json.JSONDecodeError, TypeError):
+                malformed += 1
+                continue
+            if not isinstance(record, dict):
+                malformed += 1
+                continue
+            if record.get("catalog_id") != catalog_id:
+                continue
+            action = _safe_label(record.get("action"))
+            result = _safe_result(record.get("result"))
+            if action == "evidence-delivery":
+                if result in _DELIVERY_RESULTS:
+                    hub_delivery[result] += 1
+                else:
+                    hub_delivery["invalid"] += 1
+                continue
+
+            operation_records += 1
+            actions[action] += 1
+            results[result] += 1
+            timestamp = _safe_timestamp(record.get("ts"))
+            if timestamp:
+                timestamps.append(timestamp)
+            decision = _safe_decision_ref(record.get("decision_id"))
+            if decision:
+                decisions.add(decision)
+
+            detail = record.get("detail")
+            session = detail.get("session") if isinstance(detail, dict) else None
+            if isinstance(session, dict):
+                attempted = session.get("revocation_attempted") is True
+                succeeded = session.get("revocation_succeeded") is True
+                if succeeded:
+                    cleanup["succeeded"] += 1
+                elif attempted:
+                    cleanup["failed"] += 1
+                else:
+                    cleanup["not-attempted"] += 1
+
+    timestamps.sort()
+    return LaneAuditSummary(
+        catalog_id=catalog_id,
+        operation_records=operation_records,
+        malformed_records=malformed,
+        first_ts=timestamps[0] if timestamps else "",
+        last_ts=timestamps[-1] if timestamps else "",
+        actions=dict(sorted(actions.items())),
+        results=dict(sorted(results.items())),
+        decision_refs=sorted(decisions),
+        session_cleanup=dict(sorted(cleanup.items())),
+        hub_delivery=dict(sorted(hub_delivery.items())),
+    )

src/secrets_engine/cli.py

@@ -7,12 +7,13 @@ Command surface (FR7):
     plan <decision-or-ref> --stage <stage>
     apply <decision-or-ref> --stage <stage> [--dry-run] [--bootstrap-token-file F]
     provision <catalog-id> --stage <stage> (--from-file F | --generate) --field NAME
-    verify <catalog-id> [--positive] [--negative] [--field NAME]
+    verify <catalog-id> [--positive] [--negative] [--field NAME] [--negative-token-file F]
     handoff <catalog-id> --stage <stage> --role-id-file F --secret-id-file F
     exec --catalog <catalog-id> [--field NAME] [--mode auto|npm-config|exec-env] -- CMD...
     route <catalog-id> [--json]
     revoke <catalog-id>
     lifecycle suspend|deactivate|destroy <catalog-id>
+    audit <catalog-id> [--json]
 
 Every privileged action is decision-gated and writes non-secret evidence.
 `plan` and `apply --dry-run` never mutate OpenBao.
@@ -221,6 +222,13 @@ def cmd_verify(cfg: Config, args) -> int:
         raise VerificationError(f"lane '{entry.id}' has no field to verify")
     positive = args.positive or not args.negative
     negative = args.negative or not args.positive
+    unrelated_token = None
+    if entry.stores_kv_value() and negative and args.negative_token_file:
+        from secrets_engine.openbao import read_strict_token_file
+
+        unrelated_token = read_strict_token_file(
+            Path(args.negative_token_file), purpose="negative verification token"
+        )
     if entry.stores_kv_value():
         results = []
         if positive:
@@ -234,7 +242,12 @@ def cmd_verify(cfg: Config, args) -> int:
             # Denial is path-scoped, so one probe covers every field on this path.
             results.extend(
                 run_verification(
-                    client, entry, fields[0], positive=False, negative=True
+                    client,
+                    entry,
+                    fields[0],
+                    positive=False,
+                    negative=True,
+                    unrelated_token=unrelated_token,
                 )
             )
     else:
@@ -460,6 +473,21 @@ def cmd_lifecycle(cfg: Config, args) -> int:
     return 0
 
 
+def cmd_audit(cfg: Config, args) -> int:
+    """Summarize allowlisted non-secret evidence for one cataloged lane."""
+    import json
+
+    from secrets_engine.audit import summarize_lane_evidence
+
+    entry = get_entry(cfg.catalog_dir, args.catalog_id)
+    summary = summarize_lane_evidence(cfg.evidence_dir, entry.id)
+    if args.json:
+        print(json.dumps(summary.to_json(), indent=2, sort_keys=True))
+    else:
+        print(summary.render())
+    return 0
+
+
 # -- parser ----------------------------------------------------------------
 
 
@@ -512,6 +540,11 @@ def build_parser() -> argparse.ArgumentParser:
     ve.add_argument("--field", default=None)
     ve.add_argument("--positive", action="store_true")
     ve.add_argument("--negative", action="store_true")
+    ve.add_argument(
+        "--negative-token-file",
+        default=None,
+        help="mode-0600 out-of-repo token for a real unrelated identity",
+    )
     add_token_arg(ve)
     ve.set_defaults(func=cmd_verify)
 
@@ -571,6 +604,11 @@ def build_parser() -> argparse.ArgumentParser:
         add_token_arg(lp)
         lp.set_defaults(func=cmd_lifecycle)
 
+    au = sub.add_parser("audit", help="summarize non-secret local lane evidence")
+    au.add_argument("catalog_id")
+    au.add_argument("--json", action="store_true")
+    au.set_defaults(func=cmd_audit)
+
     return p
 
 

src/secrets_engine/evidence.py

@@ -10,6 +10,7 @@ import json
 import os
 import urllib.error
 import urllib.request
+import uuid
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from pathlib import Path
@@ -55,6 +56,12 @@ class EvidenceWriter:
         day = datetime.now(timezone.utc).strftime("%Y-%m-%d")
         return self.evidence_dir / f"evidence-{day}.jsonl"
 
+    def _append_local(self, record: dict[str, Any]) -> None:
+        """Append one record before any best-effort external delivery."""
+        path = self._log_path()
+        with path.open("a", encoding="utf-8") as fh:
+            fh.write(json.dumps(record, sort_keys=True) + "\n")
+
     def record(
         self,
         action: str,
@@ -67,7 +74,10 @@ class EvidenceWriter:
         hub: bool = True,
     ) -> dict[str, Any]:
         """Append one non-secret evidence record. Returns the stored record."""
+        record_id = str(uuid.uuid4())
+        hub_requested = bool(hub and self.hub_url)
         record = {
+            "record_id": record_id,
             "ts": datetime.now(timezone.utc).isoformat(),
             "action": action,
             "result": result,
@@ -76,20 +86,38 @@ class EvidenceWriter:
             "stage": stage,
             "decision_id": decision_id,
             "detail": _scrub(detail or {}),
+            "hub_delivery_requested": hub_requested,
         }
-        path = self._log_path()
+        self._append_local(record)
-        with path.open("a", encoding="utf-8") as fh:
+        if hub_requested:
-            fh.write(json.dumps(record, sort_keys=True) + "\n")
+            delivery_result = self._post_hub(
-        if hub and self.hub_url:
+                action, result, catalog_id, stage, decision_id
-            self._post_hub(action, result, catalog_id, stage, decision_id)
+            )
+            # Append-only companion evidence makes an unavailable State Hub
+            # visible without rewriting or delaying the primary local record.
+            self._append_local(
+                {
+                    "record_id": str(uuid.uuid4()),
+                    "related_record_id": record_id,
+                    "ts": datetime.now(timezone.utc).isoformat(),
+                    "action": "evidence-delivery",
+                    "result": delivery_result,
+                    "actor": self.actor,
+                    "catalog_id": catalog_id,
+                    "stage": stage,
+                    "decision_id": decision_id,
+                    "detail": {},
+                    "hub_delivery_requested": False,
+                }
+            )
         return record
 
     def _post_hub(
         self, action: str, result: str, catalog_id: str, stage: str, decision_id: str
-    ) -> None:
+    ) -> str:
-        """Best-effort progress note to State Hub. Never raises; never sends values."""
+        """Best-effort progress note; return a non-secret delivery outcome."""
         if not self.topic_id:
-            return
+            return "skipped-no-topic"
         summary = f"secrets-engine {action}: {result}"
         if catalog_id:
             summary += f" [{catalog_id}{'/' + stage if stage else ''}]"
@@ -111,6 +139,7 @@ class EvidenceWriter:
                 method="POST",
             )
             urllib.request.urlopen(req, timeout=3).read()
+            return "delivered"
         except (urllib.error.URLError, OSError, ValueError):
             # Hub being offline must never block secret work or leak anything.
-            pass
+            return "failed"

src/secrets_engine/openbao.py

@@ -28,29 +28,34 @@ from secrets_engine.errors import BackendError, ProvisioningError
 from secrets_engine.safe_paths import containing_git_worktree
 
 
-def _check_token_file(path: Path) -> str:
+def read_strict_token_file(path: Path, *, purpose: str = "token") -> str:
-    """Read a bootstrap token file after enforcing mode-0600 and out-of-repo."""
+    """Read token material only from a mode-0600 path outside Git worktrees."""
     if not path.exists():
-        raise ProvisioningError(f"bootstrap token file not found: {path}")
+        raise ProvisioningError(f"{purpose} file not found: {path}")
     st = path.stat()
     if st.st_mode & 0o077:
         raise ProvisioningError(
-            f"bootstrap token file {path} is group/other-accessible "
+            f"{purpose} file {path} is group/other-accessible "
             f"(mode {oct(st.st_mode & 0o777)}); must be 0600"
         )
     # Refuse a token file living inside a Git worktree.
     worktree = containing_git_worktree(path)
     if worktree is not None:
         raise ProvisioningError(
-            f"bootstrap token file {path} is inside a Git worktree ({worktree}); "
+            f"{purpose} file {path} is inside a Git worktree ({worktree}); "
             "store it outside any repo"
         )
     token = path.read_text(encoding="utf-8").strip()
     if not token:
-        raise ProvisioningError(f"bootstrap token file {path} is empty")
+        raise ProvisioningError(f"{purpose} file {path} is empty")
     return token
 
 
+def _check_token_file(path: Path) -> str:
+    """Compatibility wrapper for bootstrap authentication input."""
+    return read_strict_token_file(path, purpose="bootstrap token")
+
+
 @dataclass
 class ScopedTokenSession:
     """One AppRole login token that revokes itself on close."""

src/secrets_engine/routing.py

@@ -86,7 +86,8 @@ def route_lane(
         if entry.kind == "kv" and entry.delivery_auth_management == "existing":
             missing = "externally managed OpenBao policy/AppRole readiness"
             next_command = (
-                f"secrets-engine verify {entry.id} --positive --negative"
+                f"secrets-engine verify {entry.id} --positive --negative "
+                "--negative-token-file <unrelated-token-path>"
             )
         elif entry.kind == "kv" and not entry.has_delivery_auth:
             missing = "native delivery auth declaration"
@@ -114,7 +115,8 @@ def route_lane(
             next_command = f"secrets-engine exec --catalog {entry.id} -- <command...>"
         else:
             next_command = (
-                f"secrets-engine verify {entry.id} --positive --negative"
+                f"secrets-engine verify {entry.id} --positive --negative "
+                "--negative-token-file <unrelated-token-path>"
             )
 
     return RouteResult(

src/secrets_engine/verify.py

@@ -74,10 +74,25 @@ def verify_positive(client: OpenBaoClient, entry: CatalogEntry, field: str) -> V
     )
 
 
-def verify_negative(client: OpenBaoClient, entry: CatalogEntry) -> VerifyResult:
+def verify_negative(
-    """An unrelated token must be denied. Uses an empty (invalid) token."""
+    client: OpenBaoClient,
-    # An empty/garbage token stands in for an unrelated consumer.
+    entry: CatalogEntry,
-    denied = not client.kv_can_read(entry.mount, entry.path, token="se-unrelated-denied")
+    *,
+    unrelated_token: str | None,
+) -> VerifyResult:
+    """A real unrelated token must be denied the cataloged path."""
+    if not unrelated_token:
+        return VerifyResult(
+            "negative",
+            False,
+            {
+                "reason": "no real unrelated token supplied; denial not proven",
+                "path": entry.path,
+            },
+        )
+    denied = not client.kv_can_read(
+        entry.mount, entry.path, token=unrelated_token
+    )
     return VerifyResult(
         "negative",
         denied,
@@ -167,7 +182,13 @@ def verify_auth_capability_negative(client: OpenBaoClient, entry: CatalogEntry)
 
 
 def run_verification(
-    client: OpenBaoClient, entry: CatalogEntry, field: str, *, positive: bool, negative: bool
+    client: OpenBaoClient,
+    entry: CatalogEntry,
+    field: str,
+    *,
+    positive: bool,
+    negative: bool,
+    unrelated_token: str | None = None,
 ) -> list[VerifyResult]:
     if entry.kind == "kv" and field not in entry.fields:
         raise VerificationError(
@@ -183,7 +204,9 @@ def run_verification(
         if positive:
             results.append(verify_positive(client, entry, field))
         if negative:
-            results.append(verify_negative(client, entry))
+            results.append(
+                verify_negative(client, entry, unrelated_token=unrelated_token)
+            )
     if not results:
         raise VerificationError("no verification check selected (use --positive/--negative)")
     return results

tests/test_audit.py

@@ -0,0 +1,102 @@
+import json
+import urllib.error
+from pathlib import Path
+from types import SimpleNamespace
+
+from secrets_engine.audit import summarize_lane_evidence
+from secrets_engine.cli import build_parser
+from secrets_engine.evidence import EvidenceWriter
+
+
+DECISION_ID = "e6381a56-3e55-4fac-b22c-63ee1c152ce8"
+
+
+def test_lane_audit_summarizes_allowlisted_fields_and_cleanup(tmp_path, monkeypatch):
+    writer = EvidenceWriter(
+        evidence_dir=tmp_path,
+        hub_url="http://hub.invalid",
+        topic_id="topic-id",
+    )
+    monkeypatch.setattr(
+        "urllib.request.urlopen", lambda *_args, **_kwargs: SimpleNamespace(read=lambda: b"{}")
+    )
+    writer.record(
+        "exec",
+        result="exit-0",
+        catalog_id="test-lane",
+        stage="test",
+        decision_id=DECISION_ID,
+        detail={
+            "session": {
+                "session_handle": "safe-fingerprint",
+                "revocation_attempted": True,
+                "revocation_succeeded": True,
+            }
+        },
+    )
+
+    def offline(*_args, **_kwargs):
+        raise urllib.error.URLError("offline")
+
+    monkeypatch.setattr("urllib.request.urlopen", offline)
+    writer.record(
+        "verify",
+        result="positive-pass",
+        catalog_id="test-lane",
+        stage="test",
+        decision_id=DECISION_ID,
+        detail={
+            "session": {
+                "session_handle": "another-safe-fingerprint",
+                "revocation_attempted": True,
+                "revocation_succeeded": False,
+            }
+        },
+    )
+
+    path = next(tmp_path.glob("evidence-*.jsonl"))
+    with path.open("a", encoding="utf-8") as fh:
+        fh.write("not-json\n")
+        fh.write(
+            json.dumps(
+                {
+                    "catalog_id": "test-lane",
+                    "action": "fake-SUPER-SECRET-value",
+                    "result": "fake-SUPER-SECRET-value",
+                    "decision_id": "fake-SUPER-SECRET-value",
+                    "ts": "not-a-time",
+                    "detail": {"value": "fake-SUPER-SECRET-value"},
+                }
+            )
+            + "\n"
+        )
+
+    summary = summarize_lane_evidence(tmp_path, "test-lane")
+    rendered = json.dumps(summary.to_json()) + summary.render()
+
+    assert summary.operation_records == 3
+    assert summary.malformed_records == 1
+    assert summary.actions == {"exec": 1, "invalid-label": 1, "verify": 1}
+    assert summary.results == {
+        "exit-0": 1,
+        "invalid-label": 1,
+        "positive-pass": 1,
+    }
+    assert summary.decision_refs == [DECISION_ID]
+    assert summary.session_cleanup == {"failed": 1, "succeeded": 1}
+    assert summary.hub_delivery == {"delivered": 1, "failed": 1}
+    assert "fake-SUPER-SECRET-value" not in rendered
+    assert "safe-fingerprint" not in rendered
+
+
+def test_lane_audit_empty_directory_is_a_valid_empty_summary(tmp_path):
+    summary = summarize_lane_evidence(tmp_path, "test-lane")
+    assert summary.operation_records == 0
+    assert summary.actions == {}
+    assert summary.render().startswith("Lane audit summary for 'test-lane'")
+
+
+def test_cli_parser_exposes_audit_json_command():
+    args = build_parser().parse_args(["audit", "test-lane", "--json"])
+    assert args.catalog_id == "test-lane"
+    assert args.json is True

tests/test_integration_bao.py

@@ -90,9 +90,32 @@ def test_full_chain(bao_dev, tmp_path):
 
     pos = verify_positive(client, entry, "npm_token")
     assert pos.passed, pos.detail
-    neg = verify_negative(client, entry)
+    client.write_policy(
+        "se-unrelated-denied",
+        f'path "{entry.mount}/data/{entry.path}" {{ capabilities = ["deny"] }}\n',
+    )
+    unrelated = json.loads(
+        client._run_ok(
+            ["token", "create", "-format=json", "-policy=se-unrelated-denied"]
+        )
+    )["auth"]["client_token"]
+    neg = verify_negative(client, entry, unrelated_token=unrelated)
     assert neg.passed, neg.detail
 
+    # A real unrelated identity with accidental policy overlap must make the
+    # negative check fail; an invalid/garbage token could not detect this.
+    client.write_policy(
+        "se-unrelated-overlap",
+        f'path "{entry.mount}/data/{entry.path}" {{ capabilities = ["read"] }}\n',
+    )
+    overlapping = json.loads(
+        client._run_ok(
+            ["token", "create", "-format=json", "-policy=se-unrelated-overlap"]
+        )
+    )["auth"]["client_token"]
+    leaked = verify_negative(client, entry, unrelated_token=overlapping)
+    assert not leaked.passed, leaked.detail
+
     # exec delivery: child can resolve token via npmrc; assert via a probe script
     probe = tmp_path / "probe.sh"
     probe.write_text(

tests/test_multifield_readiness.py

@@ -83,8 +83,11 @@ def test_verify_defaults_to_every_declared_field_and_one_path_denial(
     monkeypatch.setattr(cli, "_require_lane_approval", lambda *_args: None)
     monkeypatch.setattr(cli.OpenBaoClient, "resolve", lambda *_args, **_kwargs: object())
 
-    def fake_verify(_client, _entry, field, *, positive, negative):
+    def fake_verify(
+        _client, _entry, field, *, positive, negative, unrelated_token=None
+    ):
         calls.append((field, positive, negative))
+        assert unrelated_token is None
         check = "positive" if positive else "negative"
         return [VerifyResult(check, True, {"field": field, "reason": "test"})]
 
@@ -100,6 +103,7 @@ def test_verify_defaults_to_every_declared_field_and_one_path_denial(
         field=None,
         positive=False,
         negative=False,
+        negative_token_file=None,
     )
 
     assert cli.cmd_verify(_config(tmp_path), args) == 0

tests/test_redact_evidence.py

@@ -1,4 +1,6 @@
 import json
+import urllib.error
+from types import SimpleNamespace
 
 from secrets_engine.evidence import EvidenceWriter, _scrub
 from secrets_engine.redact import looks_secret, redact_text
@@ -39,3 +41,48 @@ def test_evidence_record_has_no_value(tmp_path):
     # written to disk too
     files = list(tmp_path.glob("evidence-*.jsonl"))
     assert files and "npm_shouldnotappear123" not in files[0].read_text()
+
+
+def test_evidence_records_append_only_hub_delivery_success(tmp_path, monkeypatch):
+    monkeypatch.setattr(
+        "urllib.request.urlopen",
+        lambda *_args, **_kwargs: SimpleNamespace(read=lambda: b"{}"),
+    )
+    writer = EvidenceWriter(
+        evidence_dir=tmp_path,
+        hub_url="http://hub.invalid",
+        topic_id="topic-id",
+    )
+
+    primary = writer.record("verify", result="pass", catalog_id="lane")
+
+    lines = [
+        json.loads(line)
+        for line in next(tmp_path.glob("evidence-*.jsonl")).read_text().splitlines()
+    ]
+    assert len(lines) == 2
+    assert lines[0] == primary
+    assert lines[1]["action"] == "evidence-delivery"
+    assert lines[1]["result"] == "delivered"
+    assert lines[1]["related_record_id"] == primary["record_id"]
+
+
+def test_evidence_records_hub_failure_without_raising(tmp_path, monkeypatch):
+    def offline(*_args, **_kwargs):
+        raise urllib.error.URLError("offline")
+
+    monkeypatch.setattr("urllib.request.urlopen", offline)
+    writer = EvidenceWriter(
+        evidence_dir=tmp_path,
+        hub_url="http://hub.invalid",
+        topic_id="topic-id",
+    )
+
+    writer.record("verify", result="pass", catalog_id="lane")
+
+    lines = [
+        json.loads(line)
+        for line in next(tmp_path.glob("evidence-*.jsonl")).read_text().splitlines()
+    ]
+    assert lines[-1]["action"] == "evidence-delivery"
+    assert lines[-1]["result"] == "failed"

tests/test_verification_negative.py

@@ -0,0 +1,61 @@
+import copy
+import os
+
+import pytest
+
+from secrets_engine.catalog import validate_entry
+from secrets_engine.errors import ProvisioningError
+from secrets_engine.openbao import read_strict_token_file
+from secrets_engine.verify import verify_negative
+from tests.test_catalog import VALID
+
+
+def _entry():
+    return validate_entry(copy.deepcopy(VALID))
+
+
+class NegativeClient:
+    def __init__(self, can_read):
+        self.can_read = can_read
+        self.tokens = []
+
+    def kv_can_read(self, _mount, _path, *, token):
+        self.tokens.append(token)
+        return self.can_read
+
+
+def test_negative_check_fails_closed_without_real_unrelated_token():
+    client = NegativeClient(can_read=False)
+    result = verify_negative(client, _entry(), unrelated_token=None)
+    assert result.passed is False
+    assert "not proven" in result.detail["reason"]
+    assert client.tokens == []
+
+
+def test_negative_check_uses_real_token_and_detects_policy_overlap():
+    denied = NegativeClient(can_read=False)
+    assert verify_negative(
+        denied, _entry(), unrelated_token="test-unrelated-token"
+    ).passed
+    assert denied.tokens == ["test-unrelated-token"]
+
+    overlapping = NegativeClient(can_read=True)
+    result = verify_negative(
+        overlapping, _entry(), unrelated_token="test-overlapping-token"
+    )
+    assert result.passed is False
+    assert "LEAK RISK" in result.detail["reason"]
+
+
+def test_negative_token_file_requires_mode_0600(tmp_path):
+    path = tmp_path / "unrelated.token"
+    path.write_text("test-unrelated-token", encoding="utf-8")
+    os.chmod(path, 0o644)
+    with pytest.raises(ProvisioningError, match="must be 0600"):
+        read_strict_token_file(path, purpose="negative verification token")
+
+    os.chmod(path, 0o600)
+    assert (
+        read_strict_token_file(path, purpose="negative verification token")
+        == "test-unrelated-token"
+    )

workplans/SECRETS-WP-0007-production-lifecycle-hardening.md

@@ -290,11 +290,17 @@ priority: high
 Progress 2026-08-23. KV verification now checks every declared field by default,
 with one path-level negative probe, and route readiness reads once and requires
 all declared fields. Missing readiness evidence contains field names only. Tests
-prove a missing sibling makes readiness false. The real unrelated-identity
+prove a missing sibling makes readiness false. Evidence delivery attempts now
-negative contract, audit request correlation, durable evidence delivery, and
+receive append-only local success/failure receipts, and `audit` produces an
-lane audit summary remain outstanding.
+allowlisted per-lane summary of actions, results, decisions, session cleanup,
+and delivery outcomes without echoing arbitrary detail. The real
+unrelated-identity input is now enforced as a strict mode-0600 out-of-repo token
+file; absence fails closed, and throwaway OpenBao integration deliberately adds
+an overlapping unrelated read policy and proves the check fails. Production
+identity selection/ownership, audit request correlation, queued/replayed State
+Hub delivery, and complete privileged failure-path evidence remain outstanding.
 
-The complete repository suite passes with 103 tests after these changes,
+The complete repository suite passes with 111 tests after these changes,
 including throwaway OpenBao integration coverage.
 
 Make verification and routing truthful for multi-field and high-risk lanes: