WIX0103 on the first run: Invoke-BootstrapDownload.ps1 not found because
wix build resolves File/@Source relative to the working directory, not
the wxs location. Matches the local build documented in Package.wxs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
workflow_dispatch-only, mirroring the agent release contract: validate
the runner script (PS parse + ASCII-only, it runs as SYSTEM on every
enrolling device), wix 5 build, upload versioned + stable blob names to
the agent container (no-cache, no legacy mirror - the MSI channel never
existed there), HEAD-verify the public alias serves the fresh bytes,
Sigstore provenance attestation. MSI version stays source-controlled in
Package.wxs - bumped to 1.0.2 (runner: encoding fix + bootstrap-msi.log
rename). Customer download:
https://download.autopilotmonitor.com/agent/AutopilotMonitor-Bootstrap.msi
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The agent log directory mixed three conventions (agent_20260817.log,
bootstrap_agent.log, msi-bootstrap.log) while everything else in the
diagnostics ZIP is already kebab-case. All log files now follow
<component>-<content>.log:
agent.log (was agent_YYYYMMDD.log; segments agent-NNN.log)
bootstrap-script.log (was bootstrap_agent.log; script v2.4)
bootstrap-msi.log (was msi-bootstrap.log)
crash-<utc>.log (was crash_<utc>.log)
gather-rules-debug.log (was gather_rules_debug.log)
ime-pattern-matches.log (was ime_pattern_matches.log)
The date is dropped from the main log: the agent only lives for one
enrollment, and date rotation split the chronology at local midnight -
which can fall mid-enrollment while the clock is still on the OOBE
default timezone. The existing 50MB size rotation stays as the growth
bound. Legacy dated files are ignored by the rotation probe; crash
detection still picks up legacy crash_*.log once so an in-place
self-update cannot lose a pending crash classification.
Diag collectors glob *.log and need no change. MCP zip-map and docs
note the legacy names for archives from pre-rename agents. Settings UI
hint paths updated. MSI rebuild + bootstrap script publish (v2.4) +
MCP redeploy required to ship.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Intune LOB requires a higher ProductVersion to accept a re-upload. Also
includes the runner logging fix: PS 5.1 *>> redirection writes UTF-16LE
into the otherwise ANSI msi-bootstrap.log - route through Add-Content for
a single encoding.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Still needed: the skill extracts each MSI here (msiexec /a payload for grepping
in binaries); only decompilates go to the ime-decompiles repo and MSIs to the
ime-archive blob container.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
P2 of the 2026-08-17 IME code audit. IME log wording is an unstable
contract (audit found 9 silently dead patterns, reworded lines, quote
drift, [DO TEL] removal); the registry is IME's internal product
contract - the ESP page and Intune reporting consume it.
ImeRegistryAppStateObserver + ImeRegistryAppStateHost (always-on
observability, no DecisionEngine coupling):
- one recursive RegistryWatcher on HKLM\...\IntuneManagementExtension
covers Win32Apps (EnforcementStateMessage JSON, ExitCode, Intent),
EspTrackingWin32Apps and SideCarPolicies\StatusServiceReports;
2 s debounced snapshot-and-diff + 60 s periodic fallback tick
(poll-only degraded mode if the watcher fails to arm)
- silent startup baseline: Win32Apps keys survive re-enrollments,
pre-existing state never replays as fresh events
- registry_app_state (Info, state-change-only, 200/session cap with
one-shot cap notice)
- app_state_reconciliation (Warning, once per app, immediate upload):
a changed entry terminal for >= 90 s is compared against the
log-derived AppPackageState - divergence is the built-in
log-pattern drift alarm (registry = truth, logs = narrative)
Event types dual-registered (Constants + MCP resource-catalog + shared
manifests regenerated). OKF concept doc agent/registry-app-state-pillar.md
incl. WDP Paket-3 reuse notes. Tests: pure parse/diff/divergence logic +
tick flow via snapshot override (baseline silence, emission, dwell,
fire-once, cap). All suites green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
P0b/P1 of the 2026-08-17 IME code audit (agent release required):
- Tail AppActionProcessor.log (+ rollover glob) in ImeLogTracker and the
--run-ime-matching CLI. IME routes every [Win32App][ActionProcessor]
line there (enforcement decision engine), so IME-NO-ACTION and
IME-NOT-TARGETED-1/2 could never fire before, and GRS skips were
invisible. Live session 0fe2c562 carried 16 no-action and 64
classification lines in that file.
- New patterns (wordings verified in decompiled IME 1.97 AND 1.104):
IME-GRS-SKIP ("is in GRS. The app will not be enforced." -> Postponed;
THE canonical silent non-install reason), IME-GRS-CLEARED (early
subgraph termination, retry next check-in -> Postponed),
IME-TOKEN-FAILURE ("Failed to get AAD token ... errorCode = N").
- New action imeTokenFailure -> ime_token_failure Warning event (new
EventType: Constants + MCP resource-catalog + shared manifests
regenerated). Tracker dedups per distinct errorCode; included in
the historic-replay guard so stale lines from a previous enrollment
cannot emit a misleading warning. Complements IME-TOKEN-SUCCESS:
an auth outage previously left check-ins silently doing nothing.
Tests: adapter emission + unknown-code + tracker dedup; known-actions
catalog extended; all suites green (Functions 4262, Agent 2252, web 762).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
IME 1.104 no longer logs the [DO TEL] line whose JSON carried the
DO-reported download duration - the Download Progress card lost all
times (per-app and total).
Durations now fall back to the observed event window (first download
event -> first completion-marking event) whenever the DO-reported value
is absent, and completed items show "in <duration>" inline. Works
retroactively for already-stored sessions, no agent release required;
the DO-reported value still wins when present (older IME versions).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
When RecordImeVersionAsync reports a genuinely new IME version (~monthly),
EventIngestProcessor now also enqueues an ime-msi-archive job. The worker
(QueuePollingWorker, heartbeat, pause on ImeMsiArchivingEnabled=false)
downloads the installer and stores it write-once (If-None-Match:*) in the
permanent ime-archive container ({version}/IntuneWindowsAgent.msi +
provenance.json), then merges MsiArchiveStatus/BlobPath/Sha256/Bytes/SourceUrl
onto the ImeVersionHistory row - so every fleet-observed IME build stays
available for decompilation/diffing after Microsoft's versionless CDN moves on.
The download URL comes from agent telemetry (msiDownloadUrl enrichment) and is
treated as untrusted: https-only *.manage.microsoft.com allowlist with exact
filename match, canonical versionless fallback, no redirects, size cap
(MaxImeMsiDownloadSizeMB, default 250, preflight + mid-stream) enforced by a
SHA-256-hashing pass-through stream. Archiver never throws; the worker retries
only transient statuses (download/timeout/error), permanent rejections
complete with their Failed:* status persisted.
Admin settings ImeMsiArchivingEnabled (default true) + MaxImeMsiDownloadSizeMB
via the full 4-file web chain; archive fields on the version-history API are
GA-only (non-GA projection unchanged, redaction tests extended). Shared
manifests regenerated. Tests: archiver status branches via recording subclass
(allowlist, caps, 409 idempotency, cancellation propagation), config defaults.
Backend suite 4262 green, web tsc + vitest 762 green, solution 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The EnterpriseDesktopAppManagement CSP records every MSI it enforces under
HKLM\SOFTWARE\Microsoft\EnterpriseDesktopAppManagement\<SID>\MSI\{ProductCode},
including CurrentDownloadUrl and ProductVersion. New ImeMsiInstallSourceProbe
resolves the entry matching the IME version parsed from the IME log (normalized
ProductVersion match first, IntuneWindowsAgent.msi filename fallback) and the
adapter attaches msiDownloadUrl / msiProductCode / msiProductVersion /
msiMatchedBy to the ime_agent_version event - answering which CDN/AFD endpoint
delivered exactly that IME build. Fail-soft, cached per version, additive keys
only (no new event type). Tests: probe selection/normalization + adapter
enrichment; assembly-wide probe default keeps dev machines deterministic.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The agent-log snapshot is zipped BEFORE packaging finishes, so packaging
problems (blocked paths, silent no-match patterns, locked files, evtx
export errors) were invisible in the uploaded archive - twice this made a
missing evtx undiagnosable (sessions a11102f4/3ae7528b). Every packaging
decision is now recorded in package-manifest.txt inside the ZIP: ADDED,
NO MATCH, BLOCKED (path guard), SKIPPED (caps/reparse/no user session),
FAILED (copy), EVTX EXPORTED/ERROR/TIMEOUT, plus each configured
additional path with its expanded folder+pattern. Written unconditionally
so absence of the file identifies a pre-feature archive.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The %4-encodes-slash file naming is a convention, not a contract: the
Microsoft-Autopilot-BootstrapperAgent/BootstrapperAgentServiceLogProvider
channel writes plain BootstrapperAgentServiceLogProvider.evtx. Deriving the
channel from the file name therefore produced an invalid wevtutil target
(and the wrong advice to rename the tenant diag path cost a WDP test
cycle). The channel is now resolved by scanning registered channels for a
matching LogFilePath, with the %4 convention only as fallback. A failed
export no longer skips the file - it falls through to the raw copy, which
succeeds for inactive/archived evtx files.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Live validation against session 0fe2c562 (IME 1.104.102, first session on
the repaired pattern set) showed all report-delta patterns at zero: IME
1.104 logs the delta JSON with single quotes ({"EnforcementState":...)
where older versions logged doubled quotes ({""EnforcementState"":...).
Every literal quote in the seven delta patterns (DETECTED-INSTALLED-1,
ENFORCEMENT-SUCCESS, ERROR-DOWNLOAD, ERROR-REPORT, NOT-APPLICABLE-2/3/4)
is now matched as "{1,2} to cover both serializations.
Verified against the session's AppWorkload.log via validate-ime-patterns:
ENFORCEMENT-SUCCESS 12 (was 0), ERROR-REPORT 14 (was 0); no CRITICALs;
the 2026-08-17 repair set confirmed live (ESP-TRACK-STATUS 24,
ESP-APP-REGISTERED 12, CACHE-MISS 11 in the agent's own match log).
Also observed in 1.104, tracked separately (no pattern change possible):
[DO TEL] line removed entirely - the DO download-duration display feeds
on it; needs an agent-side replacement (DO collector enrichment).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The MDM LOB Bootstrap-MSI chain (EnterpriseDesktopAppManagement CSP ->
msiexec -> runner -> bootstrap -> --install) carries a SYSTEM token with
SeTimeZonePrivilege and SeSystemEnvironmentPrivilege REMOVED. WMI
Win32_Process.Create duplicates the caller token onto the runtime, so
every MSI-delivered agent ran degraded for its whole life: tzutil exit 5
(ACCESS_DENIED) on timezone auto-set and privilege_denied on SEC-001
firmware reads (observed 2026-08-17, sessions a11102f4/0fe2c562/3ae7528b
vs. platform-script sessions 7b072755/c4c1c2d5 on the same binary).
Install mode now probes its own token (TokenPrivilegeProbe, fail-soft to
"held"): with a full token the proven WMI-first chain is unchanged; on a
stripped token the order flips to schtasks /Run first - the Task
Scheduler service starts the task with a fresh full SYSTEM token - with
a 10s process-verification poll (schtasks only reports "queued", the
2026-04-30 silent-failure lesson), WMI as degraded-but-alive fallback,
BootTrigger defer as the final net (full token there anyway).
Also: MSI runner log encoding - PS 5.1 *>> redirection wrote UTF-16LE
into the Add-Content (ANSI) msi-bootstrap.log, garbling every bootstrap
line; script output now routes through the same Add-Content pipeline.
Runner fix needs an MSI rebuild (wix build) to ship.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Rows were plain <tr onClick> so the browser offered no context-menu entry.
Each non-interactive cell now contains an invisible absolutely-positioned
Link (prefetch off) giving a real href for right-click/middle/ctrl+click;
visuals and SPA click behavior are unchanged. Cells with own buttons or
title tooltips (status, tenantId, actions) keep the row-click fallback.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
PUT rules/ime-log-patterns/{patternId} sat on the TenantAdminOrGA tier
while the handler only checks ?global=true and writes the GLOBAL pattern
with no GA re-gate - any customer tenant admin could mutate platform-wide
patterns via direct API call. The UI gate (isGlobalAdmin && globalAdminMode)
was cosmetic only.
Moved the entry to GlobalAdminOnly next to the reseed route and added a
regression test pinning both IME pattern write routes to that tier.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Audit vs decompiled IME 1.97.107: ~55/79 patterns verified verbatim, but
several were broken or dead against the current V3 pipeline.
Fixes:
- IME-ESP-TRACK-STATUS: drop trailing anchor (line grew ' with name: <n>.')
- HS-COMPLIANCE: tolerate both 'remdiation' typo and corrected wording;
new HS-COMPLIANCE-POST for the reworded post-detection line
- IME-DETECTION-RESULT: capture DetectedDifferentVersion (ordered first)
- IME-ENFORCEMENT-SUCCESS: unpin OldValue (Error/PendingReboot/null paths)
- IME-ERROR-REPORT: allow unquoted null OldValue on first write
- Disable 11 dead patterns (emitters removed with IME's old V1/V2 pipeline;
IME-DETECTED-OLD additionally collided with IME-SKIP-INSTALL-DETECTED)
New signals (existing actions, tailed files only):
- IME-ESP-APP-REGISTERED + IME-CACHE-MISS as app-name sources (replacing
dead IME-UPDATE-NAME-1/2/3)
- IME-WINGET-ERROR (WinGet failures were invisible)
- IME-EXITCODE-MAPPED-FAILED (normal mapped-exit-code failure path)
- IME-STOP-SIGNAL (service stop != OS shutdown)
- IME-RESTART-STOP-PROCESSING (reboot-gated processing abort)
Backend seed diff now also compares Enabled + Parameters - enable/disable
flips and parameter changes previously never propagated to the table.
All exact wordings verified against decompiled source; regex smoke-tested
via validate-ime-patterns.py (engine=regex, 75/75 compiled, all new
patterns fire on synthetic lines with the exact IME format strings).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two fixes from the a11102f4 diag analysis:
- enrollment_type_detected re-implemented the old CloudAssigned*-only
detection and kept reporting "Autopilot v1 (Classic ESP)" on WDP devices
while the session row already said v2. DeviceInfoCollector now uses
EnrollmentRegistryDetector.DetectEnrollmentTypeWithRule() - the same
detector as registration - and the event names the rule that fired.
- Tenant-configured .evtx paths never made it into the diagnostics ZIP:
active event-log channels hold their file exclusively locked. The
package service now exports the channel via wevtutil epl (channel name
derived from the %4-encoded file name) and streams the temp export.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
WDP batch order is fixed (User -> SLDM/MDM apps -> Scripts -> Win32Apps),
so a WDP-policy script can never observe the app phase live, and a hard
Batch-1 failure leaves zero visibility. LOB MSIs install via OMA-DM during
the first sync - before the SLDM batch - making them the earliest delivery
channel. The MSI carries no logic: it installs one fail-soft runner script
that downloads and executes the CURRENT server-hosted
Install-AutopilotMonitor.ps1, so all real bootstrap logic stays updatable
without re-shipping the MSI. Skips via the Deployed regkey when another
channel won the race. Build: wix build Package.wxs -arch x64 (WiX 5).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Field-verified on a live WDP enrollment (session 7b072755): WDP devices
carry NO Autopilot profile, so both CloudAssigned* v2 signals are
structurally absent and every WDP session was classified v1. The reliable
marker is the Device Preparation orchestration tree
AutopilotSettings\DevicePreparation\BootstrapperAgent, whose
ExecutionContext JSON carries the WDP policyId - present on WDP, absent on
classic v1 (verified there too; classic only has ESP category VALUES named
DevicePreparation.* on the parent key, hence a subkey probe).
CloudAssigned* checks stay as defence in depth. Profile seeding and the
downgrade guards from earlier today take over once registration carries v2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Field-verified against the portal list: Intune normalizes corporate
identifiers on upload - uppercased, all non-alphanumerics stripped per
component ("Microsoft Corporation,Virtual Machine,7801-5131-..." is stored
as "MICROSOFTCORPORATION,VIRTUALMACHINE,78015131...") and the device
matched + enrolled against that normalized entry. Our exact-match against
the raw device values (and the contains() filter using the dashed serial)
could therefore never hit.
Both sides now normalize before comparison, and the contains() narrowing
uses the normalized serial. Raw-form entries created directly via Graph
still match. The not-found message shows raw AND normalized form so
rejection logs line up with the portal list.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The searchExistingIdentities ACTION requires
DeviceManagementServiceConfig.ReadWrite.All, which the app deliberately does
not hold - the feature was broken in every tenant. Listing the same data via
GET works with Read.All, which the app already has for
windowsAutopilotDeviceIdentities. No new permission, no customer re-consent.
- contains(importedDeviceIdentifier,serial) filter for server-side
narrowing (mirrors AutopilotDeviceValidator), exact case-insensitive
match client-side on manufacturerModelSerial identities only
- 400 filter rejection falls back to an unfiltered paged scan
(60 x 1000 covers the portal maximum of 10 CSVs x 5000 rows);
page-cap overrun is transient, never a cached false negative
- Graph 401/403 stays a definitive, negative-cached failure with an
actionable message (now naming Read.All)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two changes from the WDP live-test follow-up:
- Remove the serialNumber-type probe again (decision: only the documented
Windows manufacturerModelSerial identifier authorizes; a serial-type entry
is an Intune-side admin misconfiguration and must not count).
- Classify Graph 401/403 on searchExistingIdentities as DEFINITIVE with an
actionable message naming the missing permission
(DeviceManagementServiceConfig.ReadWrite.All / admin consent) and
negative-cache it. Field case: the missing consent was classified
transient and trapped the agent in an endless 503 Retry-After loop.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Docs research showed the serial-only corporate identifier type is NOT the
documented WDP path (Windows only supports manufacturer,model,serial; Intune
ignores serial-type for Windows ownership). The dual probe stays because the
serial type is the only one the portal offers for MANUAL entry, so admins
land there by mistake and an uploaded serial still expresses authorization
intent. Comments and test docs now state that rationale instead of the
incorrect "WDP guidance uses serial type" claim. No behavior change.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
First live WDP enrollment 403d at device validation. Root causes:
- CorporateIdentifierValidator only searched the manufacturerModelSerial
identifier type, but the Device Preparation guidance has admins upload
SERIAL-type corporate identifiers - those never matched. The Graph
searchExistingIdentities call now probes both types in one round-trip.
- SecurityValidator overwrote deviceValidationError per validator, so the
403 Details only showed the LAST failure (CloudPc) and hid the corporate
identifier miss. Errors now aggregate in chain order and the transient
flag is sticky-OR (any transient failure yields 503 Retry-After instead
of a definitive 403).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New SessionSweepFunction (cron 0 30 * * * *, offset from the 2h Maintenance
timer at minute 0) runs only MarkStalledSessionsAsTimedOutAsync, halving the
sweep-induced classification latency; thresholds unchanged. The sweep now
returns a SessionSweepResult (counts + catastrophic-failure message) so the
new timer emits SessionSweepCompleted heartbeats / SessionSweepFailed errors
(dual-registered in the ops-alert web catalog). No lease: the sweep is
idempotent, overlap with the 2h chain is a no-op.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
From the 3b7ba31f (succeeded) vs gardner/esa (parked) hybrid session comparison:
explanation now covers the cached-credentials/pre-logon-VPN variant and why the
Hello wizard never appearing is downstream of the missing PRT; new remediation
block for VPN machine tunnels, cloud Kerberos trust, explicit Hello disable,
update deferral and pre-provisioning. Links the new docs article.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two low-risk fixes ahead of the first live Device Preparation enrollment:
- StoreSessionAsync: EnrollmentType is now sticky-v2 across re-registration
(same contract as the sticky-true booleans) so a degraded AutopilotSettings
read after a reboot can no longer flip an observed v2 session back to v1.
- EnrollmentScenarioProfileUpdater: ImeUserSessionCompleted,
SelfDeployingDeadlineConfirmed and WhiteGloveSealingConfirmed no longer
overwrite an established DevicePreparation profile (seeded at Medium from
the registry-deterministic v2 fact).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Telemetry follow-ups from the 7d tool-quality review:
- tool_call log lines now carry a capped errorMessage (thrown message or
leading toolError text) so error drilldowns name the failure class
instead of guessing from args; absent on success.
- get_rule_stats omits per-rule daily trend rows by default (they drove
50% of calls over the response cap at platform scope); new
includeTrends arg opts back in, trendsNote marks the omission.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The presence page header was a non-wrapping `flex justify-between`, so on a
phone viewport the title column was compressed to a few characters wide and
"Active Users" plus its description broke one or two words per line next to
the Window/Refresh controls.
Stack the title and controls vertically below the `sm` breakpoint and keep
the existing side-by-side row from `sm` up. The controls also wrap among
themselves on very narrow screens.
The user table had the same problem: its card used `overflow-hidden`, which
clipped the Tenant and Last seen columns off-screen with no way to reach
them. Wrap the table in the repo's usual `overflow-x-auto` container so it
scrolls inside its card, keep cells on one line, and trim cell padding to
`px-4` on mobile.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BwjVj2PvpVRk7QUjQf3i8C
A single corrupt session average (request spanning sleep/hibernate counts
wall-clock hours) inflated the request-weighted location mean by orders of
magnitude (Amersfoort: 73286 ms from one 1.38M-ms session). Add
MedianApiLatencyMs per location + global (upper median over per-session
averages, duration convention), base ApiLatencyVsGlobalPct on medians, and
switch web display/sort to the median with the weighted avg in the tooltip.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The test-only fetch stub matched the Entra authority via substring
(includes), which CodeQL flags as incomplete URL sanitization. Parse the
URL and compare the hostname exactly instead — also strictly more precise
for what the stub intends (intercept only the Entra host).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- useProgressSearch calls the server-side lookup; the client-side filter
over the tenant list is gone with the list itself.
- useProgressEvents passes the serial on every events read and refreshes
the session summary via the lookup (guarded by sessionId so a newer
re-enrollment of the same serial cannot swap the selected session).
- SignalRContext.joinGroup takes {serialNumber, onDenied}; the serial is
remembered per group and re-presented on the automatic post-reconnect
rejoin — without that, a roleless viewer's live updates would die on
the first network blip. A refused join is now visible (console.warn +
progress-page notification) instead of a silently frozen page.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Security finding: GET progress/sessions shipped the tenant's latest 100
sessions to any authenticated (roleless) user with the serial filter
applied only in the browser; events and SignalR session-group joins were
open to any known sessionId.
Model now: knowing the device's serial number is the authorization proof.
- progress/sessions/lookup (AuthenticatedUserWithRole + QueryParam)
resolves at most ONE session server-side: roleless callers need the
exact serial/device name, members/GA keep substring search. The
tenant-wide list route is gone (catalog fail-closes it).
- progress events require ?serial= re-proof on every read; missing
session and serial mismatch answer the same 404 (no existence oracle).
- Session-group joins by roleless same-tenant callers require the serial
(SerialKnowledgeProof, shared with the events route); member/GA/
delegated joins are unchanged. MemberRead on progress routes stays
deliberately rejected (regression history c4dabeee).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- /oauth/token: AbortSignal.timeout(15s) on the outbound Entra fetch;
TimeoutError maps to 504 token_exchange_timeout (502 stays for hard
failures). Undici defaults previously allowed ~300s per stalled request.
- Boot guards (fail-fast, matching config.ts posture): AUTOPILOT_ENTRA_AUTHORITY
must be https (sovereign clouds allowed, host not pinned); MCP_PUBLIC_URL pin
must be https (http tolerated for loopback dev only), trailing slashes
stripped; production log hint when AUTOPILOT_ENTRA_CLIENT_SECRET is unset
(public-client vs AADSTS7000218 misconfig).
- State one-time-use: documented as accepted residual risk (a nonce store would
reintroduce the shared state the stateless multi-replica design removed;
impact bounded by single-use PKCE-bound Entra codes + callback re-validation).
- Tests: oauth-boot-guards.test.ts (validators) + token-timeout route tests
with Entra-host-scoped fetch stub. Suite: 559 passed, tsc clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Full verify pass over all 54 built-in rules (validate_rule sweep, 30d/7d
get_rule_stats, test_analyze_rule spot checks). Fixes:
- ANALYZE-APP-007 v1.1.0: declare enrollment_failed as a non-required
condition — the exists-factor referenced an undeclared signal and could
never apply.
- ANALYZE-APP-014 v1.1.0: add non-required event_data conditions for
espCategory/espErrorCode/topCandidatePackage/candidateCount/confidence
so all explanation {{tokens}} resolve instead of rendering literally.
- ANALYZE-ENRL-002 v1.2.0: capture session_timeout.timeoutHours so
{{timeoutHours}} interpolates (verified via dry-run: renders "5").
- RuleEngine TryFlatLookup/ResolveDotPath: flatten IEnumerable data
values to a comma-joined string — List fields (e.g. app_tracking_summary
.pendingNames) stringified to "System.Collections.Generic.List`1[…]",
breaking contains-matching and {{pendingNames}} interpolation.
- GetRuleStatsAsync: raise the 500-row cap to 10000 and log when hit.
The scan walks dates ascending, so the old cap silently dropped the
NEWEST dates of a 30d window — newly added rules (e.g. ANALYZE-SEC-008)
were entirely missing from global stats.
- MCP validate_rule lint: accept event_data_array itemField as a {{token}}
resolver — {{identity}} in ANALYZE-SEC-005/006 was a false positive.
Backend 4157 tests green, MCP 548 green + tsc clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The evaluateOn feature was API/MCP-only — portal users could not see or set
evaluation triggers. The rule editor gains an Evaluation Triggers section
(enrollment end / WhiteGlove seal checkboxes + on_event type list) with
client-side grammar + hard-block validation sourced from the baked guardrails
mirror (combine.js now emits BLOCKED_INTERIM_TRIGGER_EVENT_TYPES for the web);
rule cards show an 'Interim: …' badge for interim-enabled rules (incl. the
built-in ANALYZE-ID-004). Create/edit payloads round-trip evaluateOn; the
template-copy path already deep-clones it server-side.
Suites: web 762 + tsc + lint clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A warning does not stop an end user — telemetry-cadence types (performance_snapshot,
agent_metrics_snapshot, download_progress, network_state_change/connectivity_check,
log_entry, agent_trace, stall_probe_check) would turn every ingest batch into an
analyze run. Now enforced in three layers:
- AnalyzeRuleTriggers.BlockedOnEventTypes (code-only enforcement): CRUD write is
rejected with an explicit message, and OnEventTypes() drops blocked types at
runtime so even a persisted row (pre-validation data, direct table write) is
inert for the trigger registry and the engine alike.
- rules/guardrails.json gains blockedInterimTriggerEventTypes as the display/
pre-flight mirror (baked into the MCP RULE_GUARDRAILS via combine.js); a
parity test pins mirror and code set against drift — same pattern as the
gather hard blocks.
- MCP validate_rule: the high-frequency finding is now an ERROR sourced from the
baked mirror, not a hardcoded warning.
Suites: Backend 4155, MCP 548.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Oliver Kieselbach