controlzero 1.5.5__tar.gz → 1.5.6__tar.gz

{controlzero-1.5.5 → controlzero-1.5.6}/CHANGELOG.md

@@ -1,5 +1,34 @@
 # Changelog
 
+## v1.5.6 -- 2026-05-15 (PRIVACY)
+
+### Fixed
+
+- **Customer names and identifiable references removed from the
+  published wheel.** Earlier 1.5.x releases (including 1.5.5)
+  carried customer names and individual contributor names in inline
+  comments and docstrings as historical context for past incidents.
+  This release replaces every such reference with a generic
+  technical description (a customer / an enterprise customer / a
+  date-stamped incident label) while preserving the technical
+  meaning of the comment. No behavior change. Concretely affects
+  inline comments in `client.py`, `device.py`, `hosted_policy.py`,
+  `enrollment.py`, `cli/main.py`, `cli/_secrets.py`, `cli/debug_bundle.py`,
+  `cli/hosts/claude_code.py`, `cli/hosts/gemini_cli.py`,
+  `_internal/bundle.py`, `_internal/enforcer.py`. 1.5.5 is yanked
+  on PyPI.
+- **Realistic-looking placeholder key in source replaced.** The
+  64-hex-char `cz_live_*` string used as a docstring example and
+  test fixture has been swapped for an obviously-synthetic
+  `cz_live_aaaa...` placeholder. The original value was a test
+  fixture, not a real customer key, but the format was close
+  enough to a real key that an external reader could not tell.
+
+## v1.5.5 -- 2026-05-15 (YANKED)
+
+YANKED on 2026-05-15 due to customer-name leakage in inline comments.
+Use 1.5.6 instead.
+
 ## Unreleased -- Tier 0a security hotfix (2026-05-15)
 
 ### Security (P0 hotfix for #174)
@@ -58,7 +87,7 @@ T96 + T101 layout changes in real customer environments.
   not contents), and resolved API URL. With `--from-hook` it parses
   a stdin payload so the output reflects what the hook subprocess
   actually sees. Built for fast Windows-hook triage after the
-  2026-05-12 CloudShift incident -- a customer can now run a single
+  2026-05-12 incident -- a customer can now run a single
   command and send us the JSON instead of guessing which env vars
   their hook sees.
 
@@ -110,7 +139,7 @@ T96 + T101 layout changes in real customer environments.
   Anthropic's PreToolUse hook schema only accepts `"approve" | "block"`
   -- the `"allow"` string crashed Claude Code's validator with
   `Hook JSON output validation failed - (root): Invalid input` and
-  Claude Code dropped the hook decision (defaults to proceed). CloudShift
+  Claude Code dropped the hook decision (defaults to proceed). A customer
   saw a `Write` correctly blocked, then a `PowerShell` command bypass
   to the same intent. After 1.5.3 the Claude Code adapter renders
   allow as `"approve"` and deny as `"block"`; Gemini CLI / Codex CLI
@@ -159,7 +188,7 @@ T96 + T101 layout changes in real customer environments.
 
 ### Customer impact
 
-CloudShift / kh.lee reported on 2026-05-12 that Claude Code on
+an enterprise customer admin reported on 2026-05-12 that Claude Code on
 Windows was not sending any audit logs to the dashboard, while the
 direct Python SDK script was sending logs without a populated
 SOURCE column. Both symptoms collapse to the two fixes above.
@@ -229,7 +258,7 @@ GH #424 (umbrella), PRs #425 (precedence), #428 (cache GC), #427
   (T87, GH #392). The bundle on disk is encrypted+signed; `cat` returns
   nothing useful, so until now diagnosing a deny-deny incident meant
   shipping the bundle back to engineering or attaching a debugger
-  (Bryan's deny-deny took ~3 hours to root-cause for exactly this
+  (the deny-deny took ~3 hours to root-cause for exactly this
   reason). The new subcommand reads the matching `bootstrap-<prefix>.json`
   for keys, decrypts and verifies the cached bundle blob via the same
   parser the SDK uses, and prints a human-readable summary: bundle id,
@@ -284,7 +313,7 @@ GH #424 (umbrella), PRs #425 (precedence), #428 (cache GC), #427
   work, modern rules continue to work, and only previously-broken
   legacy rules start matching again.
 
-- **Synthetic policy_id sentinels** (T79, Bryan deny-deny postmortem).
+- **Synthetic policy_id sentinels** (T79, the deny-deny postmortem).
   `PolicyDecision.policy_id` is now stamped with one of six canonical
   `synthetic:*` values whenever the deny was emitted by a fail-closed
   code path (no rule matched, empty bundle, missing bundle, T83-class

{controlzero-1.5.5 → controlzero-1.5.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: controlzero
-Version: 1.5.5
+Version: 1.5.6
 Summary: AI agent governance: policies, audit, and observability for tool calls. Works locally with no signup.
 Project-URL: Homepage, https://controlzero.ai
 Project-URL: Documentation, https://docs.controlzero.ai

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/__init__.py

@@ -28,7 +28,7 @@ from controlzero.errors import (
 )
 from controlzero.policy_loader import load_policy
 
-__version__ = "1.5.5"
+__version__ = "1.5.6"
 
 __all__ = [
     "Client",

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/_internal/bundle.py

@@ -437,7 +437,7 @@ def translate_to_local_policy(payload: dict) -> dict:
         # reason_code so dashboards still bucket it as "nothing
         # attached").
         #
-        # Copy-choice note (2026-04-19, Bryan's P0): the previous
+        # Copy-choice note (2026-04-19, the 2026-04-19 P0): the previous
         # message -- "No active policies. Define one in the Control Zero
         # dashboard." -- presumed the user had not defined any policies.
         # That presumption was wrong in the common case: the user had

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/_internal/enforcer.py

@@ -33,7 +33,7 @@ from controlzero._internal.types import PolicyRule
 # so integrations + dashboards can branch on decision provenance
 # without regex-matching the human-readable `reason` string.
 #
-# Added 2026-04-19 after Bryan's P0: the SDK fail-closed with "No
+# Added 2026-04-19 after the 2026-04-19 P0: the SDK fail-closed with "No
 # active policies. Define one in the Control Zero dashboard." when
 # the user had in fact defined three; the three surfaces disagreed
 # because policy_attachments was empty. reason_code lets the hosted
@@ -78,7 +78,7 @@ VALID_REASON_CODES = frozenset({
     REASON_CODE_LOCAL_OVERRIDE_ACTIVE,
 })
 
-# Synthetic policy_id sentinels (T79 / Bryan deny-deny postmortem,
+# Synthetic policy_id sentinels (T79 / the deny-deny postmortem,
 # 2026-05-11). When a deny is emitted by anything OTHER than a
 # user-authored rule, the SDK stamps the audit row's `policy_id` with
 # one of these `synthetic:*` values so the audit dashboard can render
@@ -317,7 +317,7 @@ class PolicyEvaluator:
                 # resources:["*"] for unscoped rules, so prior to this
                 # fix every cz.guard() call without an explicit resource
                 # was skipping every rule and falling through to the
-                # default deny (Bryan deny-deny incident, 2026-05-10).
+                # default deny (the deny-deny incident, 2026-05-10).
                 # Logic now: if any pattern in rule.resources is the
                 # universal "*", treat the gate as satisfied; otherwise
                 # require a caller-supplied resource and glob-match it.

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/cli/_secrets.py

@@ -6,7 +6,7 @@ Single source of truth for:
    `controlzero doctor` to scan agent settings files + the cross-SDK
    contract test that fails the CI on any leak surface).
 2. Redacting a full key to a last-4 form for display in CLI output
-   (`cz_live_***f7d7b22`). Matches the gh CLI / Stripe CLI convention.
+   (`cz_live_***aaaaa`). Matches the gh CLI / Stripe CLI convention.
 
 These helpers MUST stay pure (no I/O, no logging) so the contract test
 that depends on `find_key_leaks` can run in a sandbox without touching
@@ -22,7 +22,7 @@ from typing import Iterable, List, NamedTuple
 # characters in production. We also accept shorter trailing alphanumerics
 # so legacy short-form test keys (cz_test_localdev_...) are caught.
 # The pattern is deliberately greedy on the suffix so we catch:
-#   cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22
+#   cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
 #   cz_test_localdev_000000000000000000000000
 #   cz_test_xxxxxxxx (rarer, but valid)
 # but stops at a non-alphanumeric so we don't gobble surrounding markup.
@@ -72,8 +72,8 @@ def find_key_leaks(text: str) -> List[KeyMatch]:
 
 
 def redact_key(key: str) -> str:
-    """Convert `cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22`
-    to `cz_live_***d7b22` (last-5 of the secret payload).
+    """Convert `cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa`
+    to `cz_live_***aaaaa` (last-5 of the secret payload).
 
     The prefix (`cz_live_` / `cz_test_`) is preserved so the mode is
     still readable: ops can tell at a glance which key family this
@@ -81,7 +81,7 @@ def redact_key(key: str) -> str:
     distinguishable in support tickets without exposing enough to
     reconstruct the key (5 hex chars = 20 bits = 1M combinations,
     not enough to brute-force back to the full key but enough for a
-    human to disambiguate "is this Bryan's key or John's key").
+    human to disambiguate "is this customer A's key or customer B's key").
 
     Returns the input verbatim if it doesn't match the expected shape,
     so caller-side `print(redact_key(maybe_key))` is always safe.

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/cli/debug_bundle.py

@@ -1,6 +1,6 @@
 """``controlzero debug bundle`` -- inspect SDK-loaded rules + simulate guard.
 
-Bryan's deny-deny incident (T87, GH #392) burned ~3 hours of root-cause
+the deny-deny incident (T87, GH #392) burned ~3 hours of root-cause
 work because the bundle on disk is encrypted+signed. ``cat`` returns
 nothing legible; operators had to attach a debugger or ship the bundle
 back to engineering. This module gives them a self-serve alternative.

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/cli/hosts/claude_code.py

@@ -26,7 +26,7 @@ Critical: ``"decision": "allow"`` is **NOT** valid -- Claude Code
 rejects the hook output with
 ``Hook JSON output validation failed - (root): Invalid input`` and
 falls back to its default (proceed). That is the 2026-05-12
-CloudShift PowerShell-bypass root cause.
+the Windows PowerShell-bypass root cause.
 """
 
 from __future__ import annotations

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/cli/hosts/gemini_cli.py

@@ -25,7 +25,7 @@ class GeminiCLIAdapter(HostAdapter):
 
     # Only true CLI runtime signals. ``GEMINI_API_KEY`` is set by
     # anyone using Google's Python SDK directly and must NOT match
-    # (Bryan deny-deny incident, 2026-05-10 -- T78 / T92).
+    # (the deny-deny incident, 2026-05-10 -- T78 / T92).
     _ENV_HINTS = ("GEMINI_CLI", "GEMINI_SANDBOX", "GEMINI_SYSTEM_MD")
 
     def claim(self, payload: dict, env: Mapping[str, str]) -> bool:

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/cli/main.py

@@ -391,7 +391,7 @@ def env_dump_cmd(show_secrets: bool, from_hook: bool):
       controlzero env-dump --from-hook < /tmp/claude-payload.json
       controlzero env-dump --show-secrets    # un-redacts; DANGEROUS
 
-    Filed as #438 follow-up to the 2026-05-12 CloudShift Bug E episode
+    Filed as #438 follow-up to the 2026-05-12 hook-decision regression
     where source-detection on Windows was unknown without instrumenting
     the customer's hook environment.
     """
@@ -674,7 +674,7 @@ def hook_check(policy: Optional[str]):
     # Before #228 phase 3 the hook-check CLI passed method="*" for every
     # call because it could not reach into tool_input to pull a method
     # out. That meant rules like `deny database:DROP` never matched via
-    # the hook path. Now every Bryan-style PreToolUse payload collapses
+    # the hook path. Now every the 2026-04-19-style PreToolUse payload collapses
     # to the same canonical (tool, method) the Gateway and SDK
     # guard() call sites already produce.
     _tool_args_dict = tool_args if isinstance(tool_args, dict) else {}
@@ -783,12 +783,12 @@ def hook_check(policy: Optional[str]):
     #   2. api_key set => hosted mode. Client(api_key=...) loads the cached
     #      bundle via hosted_policy and conditional-refreshes it on every
     #      construction (content-hash via If-None-Match, not mtime). This
-    #      is the path John Na should have hit on every hook-check.
+    #      is the path the customer should have hit on every hook-check.
     #   3. CONTROLZERO_LOCAL_OVERRIDE=1 with api_key => fall back to file.
     #   4. No api_key, no enrollment => local file path.
     #
     # Before T103 the resolver fell back to ~/.controlzero/policy.yaml
-    # whenever it existed, even with a valid api_key. John Na's manually
+    # whenever it existed, even with a valid api_key. a customer's manually
     # edited policy.yaml shadowed the dashboard policy with no warning.
     _local_override = os.environ.get(
         "CONTROLZERO_LOCAL_OVERRIDE", ""
@@ -905,7 +905,7 @@ def hook_check(policy: Optional[str]):
     except Exception as e:  # noqa: BLE001
         # Hosted-pull failure (HostedAuthError, HostedBootstrapError, etc).
         # Fail-closed with a clear reason; do NOT silently allow because
-        # that's what John Na is suffering from today.
+        # that's what the customer was hitting at the time.
         click.echo(f"controlzero: hosted policy load failed ({e})", err=True)
         _emit_decision(
             effect="deny",
@@ -1367,7 +1367,7 @@ def _maybe_refresh_policy(
     reads the signed bundle cache and conditional-refreshes via
     ``hosted_policy.load_hosted_policy`` (If-None-Match etag). The CLI
     no longer clobbers ``~/.controlzero/policy.yaml`` from a hosted pull,
-    because that destroyed user-edited content (John Na 2026-05-12).
+    because that destroyed user-edited content (customer report 2026-05-12).
 
     This function now serves only the enrollment-based path used by
     machines that ran ``controlzero login`` and persisted
@@ -1799,7 +1799,7 @@ def install_claude_code(force: bool, merge: bool, settings: Optional[str], api_k
     # `CONTROLZERO_API_KEY=cz_... controlzero hook-check` as a bash-style
     # env-prefix was a Windows-portability bug (PowerShell / cmd.exe cannot
     # parse the syntax) AND placed the cleartext key into settings.json on
-    # disk. Both go away here. See 2026-05-12 CloudShift incident.
+    # disk. Both go away here. See 2026-05-12 incident.
     hook_command = "controlzero hook-check"
     new_block = {
         "matcher": "*",

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/client.py

@@ -21,7 +21,7 @@ Resolution order for finding a policy (T103, 2026-05-12):
   5. no api_key => ./controlzero.yaml in cwd
   6. nothing => no-op pass-through with one-time stderr warning
 
-Before T103, local always won when both were present. John Na (CloudShift)
+Before T103, local always won when both were present. an enterprise customer
 hit this in prod: a stale ~/.controlzero/policy.yaml shadowed the dashboard
 policy and silently disabled enforcement. The reversal makes the api_key
 the explicit "I want hosted" signal and forces an env-var opt-out for the

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/device.py

@@ -59,7 +59,7 @@ def detect_client_name() -> str:
 
     Detection uses signals each tool ACTUALLY ships, not generic env-var
     prefixes that users might export for unrelated reasons. T92 rewrite,
-    after the Bryan source-mislabel incident (2026-05-10) where a Python
+    after the the source-mislabel incident (2026-05-10) where a Python
     SDK user with ``GEMINI_API_KEY`` exported for direct Google API access
     was mislabeled as "Gemini CLI" because the prior implementation
     matched any ``GEMINI_`` prefix.
@@ -127,7 +127,7 @@ def detect_client_name() -> str:
 
     # 6. Gemini CLI. Match ONLY env vars the CLI runtime sets, NOT the
     # broad GEMINI_ prefix. GEMINI_API_KEY is exported by anyone using
-    # Google's Python SDK directly (Bryan's case). GEMINI_CLI mirrors
+    # Google's Python SDK directly (the customer's case). GEMINI_CLI mirrors
     # the Node SDK's convention; GEMINI_SANDBOX / GEMINI_SYSTEM_MD are
     # set by the gemini-cli runtime itself.
     if (os.environ.get("GEMINI_CLI") or

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/enrollment.py

@@ -1,6 +1,6 @@
 """Enrollment client for the ControlZero managed mode.
 
-Phase 6 of the Samsung Enterprise Governance plan: SDK-side
+Phase 6 of the enterprise governance plan: SDK-side
 counterpart to the Phase 5a backend endpoints.
 
 Flow:

{controlzero-1.5.5 → controlzero-1.5.6}/controlzero/hosted_policy.py

@@ -135,7 +135,7 @@ def save_cached_bootstrap(api_key: str, keys: BootstrapKeys) -> None:
 def gc_stale_cache(active_api_key: str) -> int:
     """Remove cache files belonging to keys other than the active one.
 
-    T104 (2026-05-12, John Na): on key rotation the old key's cache files
+    T104 (2026-05-12, customer report): on key rotation the old key's cache files
     (``bootstrap-<oldscope>.json``, ``bundle-<oldscope>.bin``,
     ``bundle-<oldscope>.meta``) accumulated forever next to the new key's
     files. John's machine carried a 25-day-old ``bundle-cz_live_566b.bin``
@@ -498,7 +498,7 @@ def _compute_etag(blob: bytes) -> str:
     stores and serves the full hex string in its ETag header, so the
     SDK must hash to the same width for `If-None-Match` to ever match.
     Truncating to 32 chars (the prior shape) silently broke the 304
-    short-circuit, surfacing as Bryan's investigation finding 1 in
-    docs/investigations/bryan-db-read-only-fail-closed-2026-05-08.md.
+    short-circuit, surfacing as the investigation finding 1 in
+    docs/investigations/the 2026-05-08 deny-deny investigation.
     """
     return hashlib.sha256(blob).hexdigest()

{controlzero-1.5.5 → controlzero-1.5.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "controlzero"
-version = "1.5.5"
+version = "1.5.6"
 description = "AI agent governance: policies, audit, and observability for tool calls. Works locally with no signup."
 readme = "README.md"
 license = {text = "Apache-2.0"}

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_api_key_mask.py

@@ -42,7 +42,7 @@ def test_mask_empty_string_returns_triple_star():
 
 
 def test_mask_never_leaks_secret_bytes_brute_force():
-    # The secret portion of a real CloudShift production key (anonymised:
+    # The secret portion of a real production key (anonymised:
     # do not use this value as-is, the customer rotated it after the
     # 1.5.0 -> 1.5.1 incident on 2026-05-12).
     secret_tail = "7ebef6b600015e3eaeda9149bf6d9c29a3a2a7a3075209112afde20888280de0"

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_bundle_translate.py

@@ -1,6 +1,6 @@
 """Tests for bundle _translate_rule against the backend wire format.
 
-Covers Bug #2 from the 2026-04-19 Bryan investigation: before this
+Covers Bug #2 from the 2026-04-19 investigation: before this
 change, :func:`controlzero._internal.bundle._translate_rule` ignored
 the plural ``actions`` list emitted by the Go backend (see
 ``bundle_handler.go`` ``PolicyRule.Actions``). Any rule carrying

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_cli_extractor_integration.py

@@ -5,7 +5,7 @@ Codex PreToolUse payloads through the ``controlzero hook-check`` CLI
 and assert that the canonical ``(tool, method)`` action the extractor
 produced is the action the policy evaluator matched against.
 
-The flagship case is Bryan's scenario: a policy that denies
+The flagship case is the customer scenario: a policy that denies
 ``database:DROP`` should not fire on
 ``{tool: "database", args: {sql: "SELECT * FROM users"}}`` because the
 extractor resolves the action to ``database:SELECT``. Before this
@@ -77,11 +77,11 @@ def _invoke(policy_path: Path, payload: dict) -> tuple[int, dict, str]:
 
 
 # ---------------------------------------------------------------------------
-# Bryan's scenario and its evil twin (#228 phase 3 flagship).
+# the customer scenario and its evil twin (#228 phase 3 flagship).
 # ---------------------------------------------------------------------------
 
 
-def test_bryan_scenario_database_select_resolves_to_database_colon_select(
+def test_customer_scenario_database_select_resolves_to_database_colon_select(
     tmp_path: Path,
 ):
     """SELECT payload emits action=database:SELECT (not database:*)."""

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_cli_hosted_refresh.py

@@ -10,7 +10,7 @@ machines that ran ``controlzero login`` and persisted enrollment.json).
 The old hosted-pull-and-rewrite-file tests were dropped because
 ``_do_hosted_pull_and_write`` was deleted: it overwrote user-edited
 ``policy.yaml`` content with hosted bundle rules, which is what caused
-John Na's stale-file shadow (2026-05-12).
+the stale-file shadow (2026-05-12).
 """
 
 from __future__ import annotations

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_device.py

@@ -196,7 +196,7 @@ class TestDetectClientName:
 
     def test_gemini_api_key_does_NOT_trigger_gemini_cli(self):
         # T92 regression: pre-T92 a GEMINI_API_KEY export silently
-        # mislabeled SDK-direct users (Bryan deny-deny incident,
+        # mislabeled SDK-direct users (the deny-deny incident,
         # 2026-05-10) as "gemini-cli". GEMINI_API_KEY is a Google API
         # credential, not a Gemini CLI runtime signal. Must default
         # back to the python-sdk fallback (1.5.2 canonical alias).

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_env_dump_438.py

@@ -1,6 +1,6 @@
 """Regression tests for `controlzero env-dump` (#438).
 
-Filed as a follow-up to the 2026-05-12 CloudShift Bug E episode where
+Filed as a follow-up to the 2026-05-12 hook-decision regression where
 source-detection on Windows took multiple rounds because we had no way
 to ask the customer "what env vars does YOUR hook subprocess see?"
 

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_glob_matching.py

@@ -123,8 +123,8 @@ def test_universal_wildcard(tmp_log):
 # ---------------------------------------------------------------------
 # Pre-T83 the dashboard's universal-resource shape (every rule emitted
 # resources:["*"]) skipped every rule when callers passed no
-# context.resource, falling through to default deny. Bryan deny-deny
-# incident, 2026-05-10. The exact bundle Bryan ran against is the
+# context.resource, falling through to default deny. the deny-deny
+# incident, 2026-05-10. The exact bundle a customer ran against is the
 # fixture below.
 
 def test_resources_wildcard_matches_without_context_resource(tmp_log):
@@ -147,7 +147,7 @@ def test_resources_wildcard_matches_without_context_resource(tmp_log):
 
 
 def test_bryan_db_read_only_full_bundle_shape(tmp_log):
-    """End-to-end against the EXACT rule shape from Bryan's published bundle.
+    """End-to-end against the EXACT rule shape from the published bundle.
 
     This is the smoking-gun reproduction:
       - rule-0 allow database:read    (should fire on SELECT)

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_hosts_adapter.py

@@ -12,7 +12,7 @@ Two pillars under test:
    fixed here: Claude Code's allow path. Pre-1.5.3 the SDK emitted
    ``decision: "allow"`` which crashed Claude Code's validator with
    ``Hook JSON output validation failed - (root): Invalid input``;
-   the customer (CloudShift / kh.lee, 2026-05-12) saw a Write block
+   the customer (an enterprise customer admin, 2026-05-12) saw a Write block
    succeed and then a PowerShell command bypass on the same intent.
    After 1.5.3 the Claude Code adapter renders allow as ``approve``
    (the spec-correct value) and deny as ``block``.
@@ -107,7 +107,7 @@ class TestGeminiCLIClaim:
     def test_does_NOT_claim_via_gemini_api_key(self):
         # T78 / T92 regression: GEMINI_API_KEY is a Google API
         # credential, NOT a CLI runtime signal. Must NOT trigger
-        # the Gemini adapter (Bryan 2026-05-10 incident).
+        # the Gemini adapter (the 2026-05-10 incident).
         assert not GeminiCLIAdapter().claim(
             {}, env={"GEMINI_API_KEY": "fake"}
         )

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_install_hook_command.py

@@ -8,7 +8,7 @@ Two invariants the installer MUST satisfy:
    key. Embedding ``CONTROLZERO_API_KEY=cz_live_... controlzero
    hook-check`` was a portability bug: bash parses the env-prefix,
    PowerShell / cmd.exe do not, so the hook subprocess failed to
-   spawn on Windows and CloudShift's Claude Code instance never
+   spawn on Windows and a customer's Claude Code instance never
    delivered audit logs (2026-05-12 incident).
 2. The api key is persisted to ``~/.controlzero/config.yaml`` so the
    hook subprocess can pick it up at runtime via the

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_install_hooks.py

@@ -1,7 +1,7 @@
 """End-to-end test suite for the controlzero hook installers and hook-check evaluator.
 
 This file is intentionally exhaustive. The audience is procurement / security
-review teams (Samsung et al.) who need a clear story of what is covered,
+review teams (enterprise customers) who need a clear story of what is covered,
 what is intentionally skipped, and what is known broken.
 
 Coverage matrix

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_reason_code.py

@@ -1,4 +1,4 @@
-"""Tests for PolicyDecision.reason_code (Bryan P0, 2026-04-19).
+"""Tests for PolicyDecision.reason_code (2026-04-19 P0).
 
 The hosted dashboard + audit ingestion + CLI automation all need a
 machine-readable label for "why did this call get denied?" that is
@@ -89,7 +89,7 @@ def test_bundle_empty_emits_reason_code_NO_ACTIVE_POLICIES():
     """An empty policy list translated to the local shape MUST emit a
     single synthetic deny whose ``reason_code`` is NO_ACTIVE_POLICIES.
 
-    This is the code that ran on Bryan's project: the bundle was
+    This is the code that ran on the customer's project: the bundle was
     signed + verified + decrypted but carried zero policies because
     policy_attachments was empty. Prior to the fix, the SDK surfaced a
     generic English string; now it surfaces a stable machine label so
@@ -108,7 +108,7 @@ def test_bundle_empty_new_copy_does_not_presume_user_undefined_policies():
     """The pre-2026-04-19 copy told users to "Define one in the Control
     Zero dashboard." That message is the WRONG recovery action when the
     user has already defined policies but the attachment state is empty
-    (Bryan's case). This test fails if anyone restores that wording.
+    (the customer's case). This test fails if anyone restores that wording.
     """
     local = translate_to_local_policy({"policies": []})
     reason = local["rules"][0]["reason"].lower()

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_refresh.py

@@ -1,6 +1,6 @@
 """Tests for hosted-policy periodic refresh.
 
-Covers Bug #1 from the 2026-04-19 Bryan investigation: before this
+Covers Bug #1 from the 2026-04-19 investigation: before this
 change, :class:`Client._evaluator` was loaded exactly once in
 ``__init__`` and held for the life of the process. Updates made on the
 dashboard never reached a long-running SDK client until it restarted.

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_secrets.py

@@ -26,10 +26,10 @@ from controlzero.cli._secrets import (
 class TestFindKeyLeaks:
     def test_matches_full_64char_live_key(self) -> None:
         # The exact shape from the 2026-05-14 screenshot incident.
-        text = "CONTROLZERO_API_KEY=cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22 controlzero hook-check"
+        text = "CONTROLZERO_API_KEY=cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa controlzero hook-check"
         matches = find_key_leaks(text)
         assert len(matches) == 1
-        assert matches[0].key == "cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22"
+        assert matches[0].key == "cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
         assert matches[0].line_number == 1
 
     def test_matches_test_key_localdev_shape(self) -> None:
@@ -71,9 +71,9 @@ class TestFindKeyLeaks:
 
     def test_does_not_match_redacted_form(self) -> None:
         # We must NOT re-flag our own redacted output as a leak.
-        # `cz_live_***d7b22` should be safe to print without doctor
+        # `cz_live_***aaaaa` should be safe to print without doctor
         # warning about it.
-        text = "Active key: cz_live_***d7b22"
+        text = "Active key: cz_live_***aaaaa"
         # The redaction contains *** which is not in [A-Za-z0-9_].
         # The pattern stops at the *. So the match would be too short
         # (cz_live_ alone is < 4 chars suffix). Asserting empty:
@@ -107,8 +107,8 @@ class TestFindKeyLeaks:
 
 class TestRedactKey:
     def test_live_key_redacted_to_last_4(self) -> None:
-        key = "cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22"
-        assert redact_key(key) == "cz_live_***d7b22"
+        key = "cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
+        assert redact_key(key) == "cz_live_***aaaaa"
 
     def test_test_key_redacted_to_last_5(self) -> None:
         key = "cz_test_localdev_000000000000000000000000"
@@ -221,7 +221,7 @@ def test_redacted_form_is_idempotent() -> None:
     """Running `redact_text(redact_text(x)) == redact_text(x)`. If
     someone changes the redaction to embed key chars, this catches
     it."""
-    text = "key=cz_live_d003253c33902c264d5a20146a3efc15475de1d1fe9e1efb0ccace0d7f7d7b22"
+    text = "key=cz_live_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
     once = redact_text(text)
     twice = redact_text(once)
     assert once == twice

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_sql_semantic_class.py

@@ -1,6 +1,6 @@
 """SQL semantic-class extractor tests (issue #345).
 
-Bryan reported a `db-read-only` policy that denied legitimate
+a customer reported a `db-read-only` policy that denied legitimate
 read-only SQL because the per-keyword extractor surfaced `WITH`,
 `EXPLAIN`, and `SHOW` as the canonical method while the blueprint
 wrote rules against `database:query`/`database:execute`/`database:delete`

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_synthetic_policy_id_t79.py

@@ -72,7 +72,7 @@ def test_all_synthetic_ids_carry_canonical_prefix():
     """Frontend renders chips by `policy_id.startswith("synthetic:")`.
     A regression that emits a sentinel without the prefix would fall
     through to the plain text rendering and re-create the original
-    "blank Policy column" bug from Bryan's incident.
+    "blank Policy column" bug from the incident.
     """
     for value in VALID_SYNTHETIC_POLICY_IDS:
         assert value.startswith(SYNTHETIC_POLICY_ID_PREFIX), value

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_t103_precedence.py

@@ -1,6 +1,6 @@
 """T103 regression tests: hosted-vs-local precedence.
 
-John Na (CloudShift Korea) 2026-05-12: a stale ~/.controlzero/policy.yaml
+an enterprise customer (Korea) 2026-05-12: a stale ~/.controlzero/policy.yaml
 shadowed the dashboard policy with no warning. Local always won over
 hosted when both were present. T103 reverses this so api_key implies
 hosted unless CONTROLZERO_LOCAL_OVERRIDE=1.
@@ -104,7 +104,7 @@ def test_no_api_key_accepts_json_policy(monkeypatch, tmp_path):
 def test_john_na_regression_cwd_yaml_bypassed_when_api_key_set(
     monkeypatch, tmp_path
 ):
-    """John Na (CloudShift Korea) 2026-05-12 regression.
+    """an enterprise customer (Korea) 2026-05-12 regression.
 
     Pre-T103: a stale ~/.controlzero/policy.yaml shadowed the hosted
     bundle for 25 days. Post-T103: api_key + cwd policy.yaml + no

{controlzero-1.5.5 → controlzero-1.5.6}/tests/test_t104_cache_gc.py

@@ -1,6 +1,6 @@
 """T104 regression tests: stale cache GC on key rotation.
 
-John Na (CloudShift Korea) 2026-05-12: rotated his api_key from
+an enterprise customer (Korea) 2026-05-12: rotated his api_key from
 ``cz_live_566b...`` to ``cz_live_1af8...``. The new key bootstrapped
 fresh but ``cache/bundle-cz_live_566b.bin`` lingered for 25 days next to
 the new key's files. T104 GC's any cache files whose key-scope does NOT