diff --git a/.claude/skills/README.md b/.claude/skills/README.md
index fe5f4f9..f5c43dd 100644
--- a/.claude/skills/README.md
+++ b/.claude/skills/README.md
@@ -12,3 +12,8 @@ are not used for grouping; group tasks inside each skill instead:
 
 The `description` in `SKILL.md` decides when Claude uses the skill. List the words users
 will actually say.
+
+Skills must work on any R1.03 system, so never use names from a particular system (items,
+sections, stations, nodes). Write placeholders such as `<Section>.<Tag>` and, where an example
+helps, use names from the sample project `CI-Eng-Sample` (`examples/CI-Eng-Sample/`).
+Test-system evidence belongs in `knowledge/progress.md` and `knowledge/issues.md`.
diff --git a/.claude/skills/system-check/SKILL.md b/.claude/skills/system-check/SKILL.md
new file mode 100644
index 0000000..9eb3042
--- /dev/null
+++ b/.claude/skills/system-check/SKILL.md
@@ -0,0 +1,89 @@
+---
+name: system-check
+description: >-
+  Read-only check of a CI Server machine, run before any CI Server work. Use when
+  the user asks which CI Server version is installed, whether CI Server is running, where it is
+  installed, whether time is synchronized, or before starting any task in another skill.
+  Also covers starting CI Server (needs administrator rights). Target version R1.03.
+version: R1.03
+source: IM 36K10J10-01EN §1.2, §1.4, §3.6, §3.7; practiced on the test system 2026-09-29/30
+status: verified (2026-09-30)
+---
+
+# System check
+
+Run this read-only check before any CI Server work. It changes nothing.
+
+## 1. Installed version
+
+PowerShell:
+
+```powershell
+$keys = @('HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*','HKLM:\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*')
+foreach ($k in $keys) { foreach ($i in (Get-ItemProperty $k -ErrorAction SilentlyContinue)) { if ($i.DisplayName -match 'CI Server') { "{0} | {1}" -f $i.DisplayName, $i.DisplayVersion } } }
+```
+
+Expected for this repo: `CI Server R1.03.00 (r10200M) | 1.03.00`. Anything else: warn the
+user that the repo targets R1.03.
+
+**From the running system** (preferred when CI Server is running): export the local node
+dataset and read `VERSION_MAJOR`, `VERSION_MINOR`, `VERSION_PACK` (1, 3, 0 = R1.03.00).
+
+```
+dssqld -e local_node.qli -d LOCAL_NODE_DF -c -l local_node.lst
+```
+
+Run it in a scratch folder; it only reads. The same row shows the node name, node number and
+licence state (`NAME`/`NODE_NAME` = `UNLICENCED` means no licence is registered).
+`dssqld -f -l dict.lst` dumps the full data dictionary (all datasets and fields) if you need
+to find another dataset.
+
+**Per-component versions** (rewritten at every start): `%TLS_ALLUSERS_PATH%\lst\tls_versions.lst`,
+one line per component: `<file>,<version>,<size>,<build date>`. On the test system all 110
+lines read `R1.03.00`; a different version on some lines means a patch is installed.
+
+**Message log**: `%TLS_LOG_PATH%\umhlog.log` holds CI Server messages
+(`--> From: <PROCESS> (<node>) on <time>` then `<FAC>-<I|W|E>-<CODE>, <text>`). Read it after
+starting CI Server to spot errors (e.g. `EQP-E-SLV_COM_BAD` = no communication with a
+station). It can contain connection strings with passwords; never copy those into the repo.
+
+## 2. Install paths (environment variables)
+
+| Variable | Typical value | Holds |
+|---|---|---|
+| `TLS_ROOT_PATH` / `TLS_ITROOT_PATH` | `C:\Program Files (x86)\Yokogawa\CIServer` | Program: `tls\com` (start scripts), `tls\exe` (tools), `tls\sup` (setup files) |
+| `TLS_ALLUSERS_PATH` | `C:\Users\Public\Yokogawa\tls` | Data root |
+| `TLS_DAT_PATH` | `…\tls\dat` | Database |
+| `TLS_HIS_PATH` | `…\tls\his` | History |
+| `TLS_SAV_PATH` | `…\tls\sav` | Saved data |
+| `TLS_LOG_PATH`, `TLS_LST_PATH` | `…\tls\log`, `…\tls\lst` | Logs and listings |
+| `TLS_WAP_PATH` | `…\tls\wap` | Web application files |
+
+`tls\com` and `tls\exe` are on `PATH`, so tools such as `dssqld` and `itmval` run from any
+prompt. Setup files (`*.sup`) are plain text (CRLF); edit them with the Setup File Editor.
+
+## 3. Is CI Server running?
+
+- Windows service `StartCIServer` ("CI Service"), start type Manual on the test system.
+- Running processes live under `…\Yokogawa\CIServer\tls\exe`. None running = CI Server stopped.
+- Most tools (`dssqld`, `itmval`, …) need CI Server running.
+- Starting it needs administrator rights (the service runs as `.\CIX_PROCESS`, binary
+  `tls\exe\ciserv.exe`). From a non-elevated shell `Start-Service StartCIServer` fails with
+  "Cannot open StartCIServer service". Start it from an elevated prompt:
+  `Start-Service StartCIServer` (or Services › CI Service › Start).
+
+## 4. Time synchronization
+
+`w32tm /query /status` and `w32tm /query /configuration` (Windows Time must be running).
+Check the source matches the design (domain controller, NTP server or `NOSYNC` on Vnet/IP).
+
+## Test system results (2026-09-29)
+
+- Version R1.03.00 (r10200M). Matches the repo.
+- Environment variables and folders as in the table; all tools and setup files named in
+  IM 36K10J10 are present except `tmtime` (a Vnet/IP station tool, not expected here).
+- CI Service stopped on 2026-09-29; started 2026-09-30 from an elevated session. About 100
+  processes then run (e.g. `ciserv`, `dssman`, `itm`, `alm`, `his`, `rptgen`, `opcexe`,
+  `durm_udp`, `eqpmdc`, `eqpmqtt`, `wfl*`).
+- `LOCAL_NODE_DF`: version 1.3.0 (R1.03.00), node 1, **UNLICENCED**, router type Endpoint.
+- Windows Time running, source `time.windows.com`; synchronized on 2026-09-30 (stratum 5).
diff --git a/.claude/skills/system-management/SKILL.md b/.claude/skills/system-management/SKILL.md
new file mode 100644
index 0000000..2d71850
--- /dev/null
+++ b/.claude/skills/system-management/SKILL.md
@@ -0,0 +1,27 @@
+---
+name: system-management
+description: >-
+  CI Server R1.03 system management and command-line tools. Use when the user wants
+  to stop CI Server, export or
+  import dataset contents (dssqld), read or write item values (itmval), load or import history,
+  check database consistency (dsschk), monitor processes/CPU/disk/network, back up or migrate
+  a system, tune or debug CI Server, or plan a system layout (standalone, separate portal, HAC,
+  host-to-host, enterprise) or time synchronization. Source IM 36K10J10-01EN.
+---
+
+# System management (R1.03)
+
+Background for this area: `knowledge/manuals/IM36K10J10-01EN.md`. Known issues:
+`knowledge/manuals/IM36K01A50-01EN.md`.
+
+Rules for every task:
+- Run the `system-check` skill first (version must be R1.03).
+- CI Server tools live in `%TLS_ROOT_PATH%\tls\exe` (on `PATH`). Run them in a scratch folder,
+  because they write output and listing files to the current directory.
+- Starting/stopping CI Server and anything that writes to the system needs an elevated
+  (administrator) session and, in this conversation, the user's chosen way of working
+  (direct vs prepare only).
+
+| Task | File | Status |
+|---|---|---|
+| Export dataset contents / dump the data dictionary | `tasks/export-dataset.md` | verified |
diff --git a/.claude/skills/system-management/tasks/export-dataset.md b/.claude/skills/system-management/tasks/export-dataset.md
new file mode 100644
index 0000000..1260317
--- /dev/null
+++ b/.claude/skills/system-management/tasks/export-dataset.md
@@ -0,0 +1,52 @@
+---
+task: Export dataset contents with dssqld
+version: R1.03
+source: IM 36K10J10-01EN §2.9.2, §5.6; practiced on the test system 2026-09-30
+status: verified
+---
+
+# Export dataset contents (dssqld)
+
+Read-only. Needs CI Server running. Works from a normal or elevated prompt.
+
+## Export a whole dataset
+
+```
+dssqld -e <output>.qli -d <DATASET> -c -l <listing>.lst
+```
+
+- `-d` dataset name (e.g. `ITEM_DF`, `STATUS_DF`, `LOCAL_NODE_DF`, `CLASS_DF`, `OBJECT_DF`).
+- `-c` comma-separated output: first line is the field names, then one quoted row per record.
+- `-l` writes the result summary to a listing file.
+- `-a` also exports hidden fields.
+
+The console (and listing) end with a summary; check `Errors` is 0:
+
+```
+Data set                   Recognized  Exported    Not found   Errors
+ITEM_DF                    184         184         0           0
+```
+
+## Find datasets and fields
+
+```
+dssqld -f -l dict.lst
+```
+
+Dumps the data dictionary: every dataset with its description and fields (name, type, length,
+description). Search it to find which dataset holds what, e.g. the version fields are in
+`LOCAL_NODE_DF`, `HOST_NODE_DF` and `ENTERPRISE_SERVER_DF`.
+
+## Useful datasets seen on R1.03
+
+| Dataset | Contents |
+|---|---|
+| `LOCAL_NODE_DF` | This node: name, number, version (`VERSION_MAJOR/MINOR/PACK`), licence state |
+| `ITEM_DF` | Item definitions: full name, section path, description, limits, format, type, station, I/O address |
+| `STATUS_DF` | Item status definitions: name, number, alarm state, priority, colour, mnemonic (e.g. 9 = NORMAL) |
+
+## Filtered exports
+
+`dssqld -e <output> <filter>.qli` exports using a filter file with `@<DATASET>`, `@FIELD`
+and `@SELECT` keywords (IM 36K10J10 §5.6). The exact filter syntax is **not yet verified**;
+export the whole dataset and filter the CSV instead until it is.
diff --git a/CLAUDE.md b/CLAUDE.md
index 8556ea5..012f36f 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -1,28 +1,49 @@
 # CI Server Engineering Assistant
 
 You act as an experienced Yokogawa CI Server engineer. Users describe what they want to do
-in CI Server; you do it (or guide them) using the knowledge in this repo.
+in CI Server; you do it (or prepare it for them) using the knowledge in this repo.
 
-- **Target version:** CI Server **R1.03**. Say so if a request depends on another release.
-- **Language:** use the English (EN) manuals only.
+## Before any CI Server work
+
+1. **Confirm the software version.** This repo covers CI Server **R1.03** (R1.03.00).
+   Check the installed version first (skill `system-check`). If it is not R1.03,
+   tell the user before doing anything and point out which content may not apply.
+2. **Ask how the user wants to work** — once per conversation, before the first action:
+   - **Direct**: you make changes in CI Server yourself (run tools, edit setup files, load data).
+   - **Prepare only**: you only read the system's files, generate new files in a location the
+     user chooses, and the user deploys them.
+
+   Remember the answer for the rest of the conversation. Reading files and read-only checks
+   are always allowed.
 
 ## Where to look, in order
 
 1. **Skills**: `.claude/skills/<area>/`. There is one skill per CI Server area (e.g. Edit Module,
    Alarm Management). `SKILL.md` is an index; open only the task file you need.
-2. **Shared knowledge**: `knowledge/`. The document map, glossary, known issues and
-   system-wide concepts.
+2. **Shared knowledge**: `knowledge/`. Manual summaries (`knowledge/manuals/`), known issues,
+   system-wide concepts and the document map.
 3. **Examples**: `examples/`. Verified, working artifacts to copy and adapt.
-4. **The manuals**: if the repo does not cover it, look it up in the EN PDFs. They are **not**
-   in this repo. Default location: `C:\Users\Public\Yokogawa\tls\doc\EN\`. Use
-   `knowledge/document-map.md` to find the right manual by IM number.
 
-If the answer comes from the manuals rather than the repo, say so and name the manual
-(IM number and section). Never guess CI Server behavior; say what you could not verify.
+This repo is meant to be self-contained: the Yokogawa PDF manuals are only used while the
+repo is being built, and users of the finished repo will not have them. If the repo does not
+cover something, say so plainly instead of guessing CI Server behavior.
 
 ## Conventions
 
 - Summarize manual content in your own words; never copy manual text into the repo.
 - Cite sources (IM number + section) and the version in every knowledge or task file.
 - Mark whether content is **verified** on a real system or **from docs** only.
+- Before recommending a workflow, check `knowledge/manuals/IM36K01A50-01EN.md` (known issues).
 - Never modify the manual folder.
+- Skills use placeholders (`<Section>.<Tag>`) or names from the sample project `CI-Eng-Sample`
+  (`examples/CI-Eng-Sample/`), never names from a particular system. Names like `AID.*` in
+  `knowledge/progress.md` and `knowledge/issues.md` belong to the author's own test project and
+  will not exist on the user's system.
+
+## Building the repo (development stage)
+
+The manuals are at `C:\Users\Public\Yokogawa\tls\doc\EN\` on the development machine. Work
+proceeds manual by manual in the order in `knowledge/progress.md`: summarize the manual into
+`knowledge/manuals/`, practice each task on the test system, and record a task in a skill only
+after the user confirms it works. Track status in `knowledge/progress.md`. Log open questions
+and problems in `knowledge/issues.md`, each with the files to update once it is resolved.
diff --git a/examples/CI-Eng-Sample/README.md b/examples/CI-Eng-Sample/README.md
new file mode 100644
index 0000000..5102468
--- /dev/null
+++ b/examples/CI-Eng-Sample/README.md
@@ -0,0 +1,36 @@
+# CI-Eng-Sample (placeholder)
+
+**Status: planned, not built yet.** Nothing in this folder can be imported yet.
+
+A small, invented sample project that anyone with this repo can load into a CI Server
+**R1.03** test system. Skills use it for their examples, and every skill task can be practiced
+and verified against the same known items on any machine.
+
+## Planned contents
+
+Built up phase by phase (see `knowledge/progress.md`):
+
+| Phase | Adds |
+|---|---|
+| 5 Items and acquisition | Top-level section, sub-sections, template items, internal items (no PLC needed), items with alarm limits |
+| 6 Classes and objects | A class and objects created from it |
+| 8 Alarms | Alarm settings on sample items |
+| 9 History | An item storage group |
+| 10 Graphics | A display and symbol |
+| 11 Reporting | A report |
+
+## Planned files
+
+| File | Purpose |
+|---|---|
+| `*.qli` | `dssqld` data files, one per dataset |
+| `load.cmd` | Imports the data files in dependency order: template items → classes (compiled) → objects → remaining items (IM 36K10J10 §5.6) |
+| `remove.cmd` | Removes the sample project from a system |
+
+## Rules
+
+- Test systems only. Loading writes to the CI Server database.
+- Invented names only; no real plant or customer data.
+- The top-level section name is still to be decided. Item names may not allow `-`, so it may
+  differ from the folder name (to be checked in phase 5).
+- Every file notes the CI Server version it was built and tested on.
diff --git a/examples/README.md b/examples/README.md
index 93405c0..4cf2895 100644
--- a/examples/README.md
+++ b/examples/README.md
@@ -2,3 +2,7 @@
 
 Working artifacts Claude can copy and adapt. Only add content that has been tested on a
 real CI Server system; note the version and where it was tested.
+
+| Folder | Contents |
+|---|---|
+| `CI-Eng-Sample/` | Sample project to import into a test system; skills use its names in examples (planned) |
diff --git a/knowledge/README.md b/knowledge/README.md
index 7cb08b0..0bfaa1f 100644
--- a/knowledge/README.md
+++ b/knowledge/README.md
@@ -5,4 +5,7 @@ that area's skill (`.claude/skills/<area>/reference/`).
 
 | File | Contents |
 |---|---|
+| `issues.md` | Issue register: open questions and problems to come back to |
+| `progress.md` | Working order of the manuals and status of every practiced task |
+| `manuals/` | One summary per manual (`IM36K….md`), written in our own words |
 | `document-map.md` | All EN manuals by IM number, with editions |
diff --git a/knowledge/issues.md b/knowledge/issues.md
new file mode 100644
index 0000000..5f2b142
--- /dev/null
+++ b/knowledge/issues.md
@@ -0,0 +1,25 @@
+# Issue register
+
+Open questions and problems found while building the repo.
+
+**Status:** open · investigating · closed
+
+Item and station names such as `AID.*` come from the author's own test project on the
+development system (see "Test system" in `progress.md`); they do not exist on other systems.
+
+## How to close an issue
+
+1. Confirm the answer with the user (practiced on the test system where possible).
+2. Update **every file in the issue's "Update on close" column** with the answer.
+3. Update the matching row in `knowledge/progress.md` (status and notes).
+4. Set the issue to **closed** and write the resolution and date in "Next step / resolution".
+   Keep closed issues in the table as history.
+
+## Issues
+
+| ID | Date | Area | Issue | Status | Next step / resolution | Update on close |
+|---|---|---|---|---|---|---|
+| ISS-001 | 2026-09-30 | ITEM (`itmval`) | Writing an item with `itmval -p -n AID.WRPS.SIM.INFLOW -v 400` added the `uoff` option flag (options 0x…8a → 0x…8e). The flag stayed after writing the original value back. Meaning of `uoff` unknown; not found in the EN manuals by text search. | open | Find what `uoff` means and how to clear it (ITEM Service IM 36K10L35/L36, Engineering Module I/O Processes IM 36K10C10). Until then, `itmval -p` stays "awaiting confirmation". | Create `.claude/skills/system-management/tasks/read-write-item.md` (with the `uoff` explanation) and add it to that skill's `SKILL.md` task table; `knowledge/progress.md` row "Write an item with `itmval -p`" |
+| ISS-002 | 2026-09-30 | DATABASE (`dsschk`) | `dsschk -d ITEM_DF -v` (check only) exits 0 with no console output and nothing new in `umhlog.log`. Where results are reported is unknown. | open | Revisit with the Data Access / Data Set Service manuals; try without `-d`, or with `-e`. | Create `.claude/skills/system-management/tasks/consistency-check.md` and add it to that skill's `SKILL.md`; `knowledge/manuals/IM36K10J10-01EN.md` §4.1.1 consistency tools (add where output goes); `knowledge/progress.md` row "Consistency check" |
+| ISS-003 | 2026-09-30 | DATABASE (`dssqld`) | Filtered export (`dssqld -e <out> <filter>.qli`) failed with a guessed filter syntax (`@ITEM_DF`, `@FIELD`, `@SELECT`). | open | Get the exact filter-file syntax from the Data Set Service manuals (IM 36K10L50 / P30). | `.claude/skills/system-management/tasks/export-dataset.md` section "Filtered exports"; `knowledge/manuals/IM36K10J10-01EN.md` §5.6 step 1; `knowledge/progress.md` row "Export a dataset" |
+| ISS-004 | 2026-09-30 | ITEM (`itmval`) | `itmval -g -a <1..5>` returns `ITM-E-WRONG_ATT` on `AID.WRPS.*` items, so reading limits is unproven. | open | Retry on an item that has alarm limits configured (phase 8, alarms). | `.claude/skills/system-management/tasks/read-write-item.md` (limits section); `knowledge/progress.md` row "Read an item with `itmval -g`" |
diff --git a/knowledge/manuals/IM36K01A50-01EN.md b/knowledge/manuals/IM36K01A50-01EN.md
new file mode 100644
index 0000000..ce18d65
--- /dev/null
+++ b/knowledge/manuals/IM36K01A50-01EN.md
@@ -0,0 +1,183 @@
+---
+manual: IM 36K01A50-01EN
+title: CI Server Release Information
+edition: 3rd Edition, Sep. 2023
+version: R1.03
+status: from docs
+---
+
+# Release Information: known issues and cautions (R1.03)
+
+Known limitations and workarounds that the other manuals do not cover. Check this list
+before recommending a workflow in any area. Section numbers refer to the manual.
+
+## Platform and installation
+
+| Area | Issue | What to do |
+|---|---|---|
+| Dual-redundant platform (1.1) | Without a Vnet/IP card, guest time sync by NTP makes host messages in the historical alarm overview show wrong timestamps. | Accept, or use Vnet/IP time sync. |
+| Dual-redundant platform (1.1) | After both computers restart, the former standby may come up as *Guest OS (Isolated)*. | In the Redundancy Management Tool set Next Status to *Standby*, then restart that computer. |
+| IT Security tool (1.3) | Remote Desktop to the CI Server may fail after applying IT Security if OS patch levels differ. | Patch both the server and the RDP client to the latest level. |
+| IT Security tool (1.3) | Changing Windows region/language after applying IT Security causes problems. | Set region and language **before** applying IT Security; never change afterwards. |
+| Windows (1.4) | Windows Server 2016 start menu may not refresh after install. | Restart after installing CI Server. |
+| Windows (1.4) | CI Server started from the command line on Win10 / Server 2016 needs elevation. | Always use *Run as administrator*. |
+| Windows (1.4) | Custom fonts can cause visual glitches in the Engineering and Edit modules. | Avoid custom fonts. |
+| Install path (1.9) | The MELSEC EZSocket installer fails if CI Server is in a non-default path that contains spaces. | Non-default install paths must have no spaces. |
+| Virtual machines (1.19) | Live migration, failover and shared-storage path switching lose acquired data/alarms; failover during engineering can corrupt the engineering data. | Migrate only when no engineering is in progress; keep engineering backups. |
+| Virtual machines (1.19) | Serial-only drivers (Siemens 3964R, telemetry integration), Siemens SAPI-S7 and MELSEC are unavailable on VMs; many other drivers work over TCP/IP only. | Check driver support before choosing a VM. |
+| Virtual machines (1.19) | Do not use VM replication; do not mix physical and virtual machines in HAC; only thin clients may RDP into a virtual CI Server. | Design accordingly. |
+| Virtual machines (1.19) | The software medium must be mounted as an ISO on the VM (e.g. via Hyper-V Manager › Media › DVD Drive › Insert Disk). | Read "DVD drive" in the manuals as the mounted ISO drive. |
+
+## Services
+
+**ACCESS (1.5)**
+- OPC classic on Win10 / Server 2016 / 2019 needs .NET 3.5. Install it from the Windows medium
+  (`DISM /Online /Enable-Feature /FeatureName:NetFx3 /All /LimitAccess /Source:<drive>:\sources\sxs`),
+  then re-run the OPC Core Components Redistributable in `%TLS_EXE_PATH%`.
+- OPC AE: items in the tree root (no section) appear as `.NAME` and some clients can't use them
+  as alarm sources. Put alarm-source items inside a section.
+
+**ALARM (1.6)**
+- Shelving cannot restrict OPC AE items by status/type.
+- Shelving more than 500 alarms can overflow the DURM queue; after a Portal–Core link loss,
+  restart the shelf overview.
+
+**BUS (1.7)**
+- Secure MDUR on Linux: create `/tls/pki` manually before configuring the secure connection.
+
+**DATABASE (1.8)**
+- With Active Directory login, up to ~15 s may pass between the login form closing and the
+  application appearing. Expected behavior.
+
+**EQUIPMENT (1.9)**
+- MELSEC iQ-R needs Mitsubishi EZSocket 4.4: uninstall older EZSocket first, then run
+  `RunInstaller` (as administrator) from `%TLS_ROOT_PATH%\tls\exe\EZSocketPcRedistributableInstaller`.
+- IEC 61850: add the line, then the station (General tab: device IP; Options tab: domain prefix
+  and report object), then points. The element address is `<domain prefix><rest>`; the point
+  I/O address is only the part **after** the prefix (the prefix is added automatically).
+  Wrong delimiting means no reports/elements are created.
+
+**PROCESS (1.10)**
+- Java classes that access datasets must get the DSS session with `getDssSession()` (never
+  `Dss.createSession()` or `DssContext.getSession()`), must never close it, and must be
+  recompiled and reset after the change. Otherwise results are undefined
+  (e.g. `ADT-E-ECL_UNKNOWN` in the system log).
+- When making items with SIGNALs, handle root-level items: if `SECTION` is empty use the tag
+  alone, else `SECTION || "." || tag`. Prefer automatic naming.
+- Section-based SIGNAL creation does not support process area definitions.
+
+**REPORT (1.11)**
+- AD + printer management can make report generation fail (looks for the user's default
+  printer). Workaround: create a printer destination not bound to a device and set it as
+  the default destination in the AD user profile.
+- Report features of Schemes take effect only after restarting the Engineering module.
+
+**USER (1.12)**
+- Remote Connect is for client machines **without** CI Server; don't install it on a CI Server.
+- HTML5 displays cannot be validated via Remote Connect; engineer them on a CI Portal.
+- HTML5 is a lightweight/mobile HMI with fewer components than Java displays (details in
+  Edit Module, IM 36K10C50). Use Google Chrome; other browsers may render differently.
+- SSO leaves a landing page open that the user must close. SSO needs Active Directory, so
+  Windows only.
+- `jwuiutl` over Remote Desktop is denied by default. To allow it: Setup File Editor ›
+  `dssana.sup` › *Login Control* tab › row *Client system name `<remote>`, Display and Symbol
+  Utility* › Login method = *No* › restart CI Server. Needed by CI Exchange (CENTUM VP /
+  Prodigy) load and conversion scripts.
+- Remote Connect operator interface may fail to start (Java network read issue): retry, or
+  clear the Java cache (Java Control Panel › Temporary Internet files › Delete files).
+- Auto-login is for server / CI Portal applications only, not Remote Connect.
+- "Application blocked" for `.jnlp` over HTTPS is caused by the self-signed web server
+  certificate. Fix with a CA/domain-issued certificate (preferred) or add
+  `https://localhost:8443` to trusted sites (less safe).
+- Web HMI display names are case sensitive.
+- Custom Java components must be signed by Yokogawa before use.
+- Edit module quirks: item properties editor Value field font too small; *Update
+  item/parameter* on the Attributes sheet ignores the first change (repeat it or use the
+  Actions sheet); a symbol custom property must not reuse an existing property name;
+  choosing *Open display* in an event's Action combo adds a duplicate event (delete it);
+  function components must stay inside the drawing area and must not be grouped.
+- Same-size labels may render at slightly different sizes (Java anti-aliasing). Use *Apply
+  text smoothing* in the CI Portal setup file, or a non-integer zoom such as 99.999 %.
+- HTML5 limitations: no login name in the status bar; frame border styles map imperfectly;
+  Windows display scaling must be 100 %; no rotation or column filters on alarm overview /
+  shelved alarm overview; transparency and margin properties are ignored.
+
+## HMI, GIS and OPC UA
+
+**HMI general (1.13)**
+- Date/Time field with *editable* or *disable* set shows a black foreground in the operator
+  interface (correct in HTML).
+- Up/down buttons differ in size between operator interface and HTML.
+- With *Fit display*, tooltips and selection dialogs of trend / alarm overview are misplaced.
+  Uncheck *Zoom enabled* on those components, or design displays at the target resolution.
+
+**GIS (1.14)**
+- Click info is not shown for symbols built with a data bar. Use simple symbols as map markers.
+- Don't use symbols containing alarm overview, trend or report components in GIS (high memory
+  in Chrome; alarm overview filters don't receive item names).
+- Symbols/map layer may disappear or draw behind the map after long runs (Ctrl+wheel zoom
+  restores). In Chrome, disabling hardware acceleration fixes symbols drawn behind the map.
+
+**OPC UA (1.15)**
+- PubSub upload fails with *Internal error* if a published dataset has zero items. Remove or
+  revert those datasets.
+- The datatype combo in PubSub Reader item mapping has no effect.
+- Variables/properties of `CIAlarmConditionType` / `CISystemEventType` nodes can't be read
+  (`BadNotReadable`) or subscribed individually (`BadNotSupported`); subscribe to the alarm event.
+- CI Core memory for external OPC UA clients: under 20,000 items no extra memory;
+  20,000–99,999 add 8 GB; 100,000–199,999 add 16 GB; beyond that add 8 GB per extra 100,000.
+
+## Server Core / Portal (1.16)
+
+- Secure DURM with an invalid certificate common name continues **unencrypted** (only logged).
+  After installing/refreshing certificates, check the message log that they were accepted.
+- Remote Connect Engineering module is slow against a redundant CI Core. Use RDP to the Core,
+  or temporarily run the redundant platform in single mode.
+- Several MQTT stations on one MQTT line with brokers unavailable can time out during offline
+  engineering. Use one station per line, or a temporary local broker.
+- GIS component properties *Apply* increases Edit module memory each time; close the dialog
+  and display to recover.
+- Object Selection dialog can hang the Edit module when a section has many objects (DSSSDA at
+  high CPU). Type the object name directly; to recover, stop that DSSSDA instance in
+  Performance Monitor, close the Edit module, restart CI Server.
+- Enterprise Edit Module: shelved alarm overview shows nothing if added before the alarm
+  overview, on another layer, or with different alarm source locations. Add the alarm overview
+  first, same layer, same source locations.
+- Alarm overview CI attributes: changing checkboxes on the Attributes tab can reorder the
+  Columns tab. Workaround: *Uncheck all* › Apply › re-check › Apply › reorder on Columns › Apply.
+- Engineering module CI attribute names: pick *Add new record* or an existing name from the
+  drop-down; cancel the *Add new CI attribute* dialog if it appears unexpectedly.
+- OPC UA alarm event type filters and alarm template settings that are in use can't be
+  deleted, and a same-named one on another station also becomes undeletable. Use unique names
+  per station; to recover, copy under a new name, repoint, then delete.
+- EALs can't be edited in bulk in the Engineering module. Use `dssqld` on dataset
+  `EXTERNAL_ALARM_DF`.
+- Retrieve from archive does not restore alarm attribute databases (alarms visible, attributes not).
+- ODBC writes: Microsoft Access can't write to R1.03.00+; SQL Server Management Studio and
+  Excel via ODBC can't either. PowerShell and custom applications can (ODBC DSN
+  `CiSimbaClient`, e.g. `UPDATE ITEM_VAL SET ITEM_VALUE_STR=..., STATUS_NUMBER=... WHERE NAME=...`).
+  Alternatives: Excel Add-in, or `dssqld` on the Core/Portal.
+- Acknowledging external alarms over ODBC: filter on `NAME` in `ALARM_CURRENT`
+  (e.g. `UPDATE ALARM_CURRENT SET ACK = 1 WHERE NAME = '<time>/<id>'`), not
+  `ITEM_NAME` / `ALARM_ORIG_TIME`.
+
+## CI Exchange tools (1.17, 1.18)
+
+- CENTUM VP and Prodigy tools use `jwuiutl`; over RDP, first allow it in `dssana.sup` (see USER).
+- HTML5 conversion warnings about graphics outside the viewing area are harmless; move Generic
+  / JavaScript function components into view and reconvert to silence them.
+- *Illegal reflective access* Java warnings during load are harmless.
+- Folders must not contain Japanese characters (project, CI Exchange project and publish
+  folders; `Load.cmd` fails with `UnicodeEncodeError`).
+- `Unknown encoding 'ms932'` from `Load.cmd`: in the publish folder run `chcp 1252`, then `load.cmd`.
+- Loading a publication overwrites VNET station settings (on scan, recover 10 s, slow 5 s,
+  medium 3 s, fast 1 s) and line device (`/dev/bk00`). Correct them manually afterwards.
+- RGS with EAL: if both Vnet/IP buses to an FCS are cut, faceplates don't show OFFLINE and keep
+  old values. Set the alarm status manually, or map the Vnet communication status to an item.
+- Unsupported controllers (e.g. GSGW, APCS) get stations created but no data. Exclude them in
+  the Import Wizard (RGS with EAL), or filter in the Publish Wizard, e.g.
+  `TYPE LIKE '%%' AND STATION <> 'FCS0256' AND STATION <> 'FCS0258'`; otherwise delete the
+  stations in the Engineering module.
+- Tuning panel SUM is shown even when CENTUM integration is disabled. To reset to 0: disable
+  integration in CENTUM, then set SUM to 0 on the CI tuning panel.
+- RGS collecting CENTUM alarms: 30 or more controllers needs +8 GB memory.
diff --git a/knowledge/manuals/IM36K10J10-01EN.md b/knowledge/manuals/IM36K10J10-01EN.md
new file mode 100644
index 0000000..42802ce
--- /dev/null
+++ b/knowledge/manuals/IM36K10J10-01EN.md
@@ -0,0 +1,263 @@
+---
+manual: IM 36K10J10-01EN
+title: CI Server System Design/Management
+edition: 3rd Edition, Sep. 2023
+version: R1.03
+status: from docs
+---
+
+# System Design / Management (R1.03)
+
+How CI Server is built, how to choose a system layout, and how to run, back up, monitor,
+tune and migrate it. Section numbers refer to the manual.
+
+## 1. Fundamentals
+
+**Three tiers (1.1)**
+- **CI View**: the client HMI (Java or HTML5), on Windows PCs, tablets and phones.
+- **CI Portal**: web server that sits between CI View and CI Core.
+- **CI Core**: data server. Talks to field devices (EQUIPMENT drivers), processes real-time
+  data, manages alarms, stores history, exposes external interfaces.
+
+**Node, services, processes (1.2)**
+- One CI Server per machine; each is a **node**. The license is tied to the node.
+- A node runs **services** (DATABASE, ITEM, ALARM, EQUIPMENT, REPORT, …). Which services start
+  depends on the license.
+- Each service is made of **processes**; a process can have several **instances** (e.g. several
+  report generators, requests shared out by priority).
+- Processes talk by **DUR messages**; each has a named queue, visible in Performance Monitor.
+
+**Data model (1.3)**
+- **Item**: the basic data point. Receives acquired values, sends operator/application writes
+  to the RTU, drives alarms, serves as class I/O and as dynamic properties of displays/symbols.
+- **Section**: hierarchical folder for items (item full name `Section1.Section2.Tag`).
+- **Object**: a set of items managed together; used by processing and linked to graphics.
+- **External alarm (EAL)**: alarm from an external system received via the OPC UA client;
+  EAL settings are kept in sections.
+- **Dataset**: table (rows/fields, like an RDB table) holding configuration, current data and
+  history. Users can add their own. Read by displays, symbols, classes, objects, reports and
+  external applications.
+
+**Two layers (1.4)**: CI Portal services and CI Core services, on the same or separate machines.
+
+| Service | CI Core | CI Portal |
+|---|:-:|:-:|
+| BUS | ✓ | ✓ |
+| DATABASE | ✓ | ✓ |
+| HISTORY, AUDIT, ITEM, EQUIPMENT, ALARM, PROCESS, REPORT, ACCESS | ✓ | |
+| USER | | ✓ |
+| INTEGRATION | ✓ | ✓ |
+
+## 2. Inside CI Server
+
+| Function | Service(s) | Summary |
+|---|---|---|
+| Data acquisition (2.1) | EQUIPMENT | Drivers for STARDOM FCX, Vnet/IP, Modbus, DNP3 and others. |
+| Alarming (2.2) | ALARM | Detection (by CI Server or the RTU/PCS, with filters against chattering); notification to display, report, OPC AE/UA, sound, e-mail, printer; state and acknowledgement; stored in `ALARM_CURRENT` (current) and `ALARM_CHRONO` (history); analysis with ASPA. |
+| Historization (2.3) | HISTORY | Storage groups made of storage units, one kind per data type: items (`ITEM_HISTORY`), alarms (`ALARM_CHRONO`), audit (`AUDIT_TRAIL`), system log (`SYS_LOG`), reports. |
+| Symbol and processing (2.4) | PROCESS | Class → object → symbol (below). |
+| Display (2.5) | USER | Displays contain symbols; appearance comes from layouts in an XML layouts file. User definitions set the initial display/layout; *Open display* actions name a display/layout. Defaults: display `main`, layout `default`. |
+| Reporting (2.6) | REPORT | Configure in the Engineering module (query, schedule, trigger item, destination) → generate by schedule, trigger item or ad hoc → output via spool statement or destination (file, printer, application) → stored in history with a lifetime set by the storage group. |
+| User authorization (2.7) | USER | Users belong to authorization groups; groups grant actions. Can be managed through a Windows domain (see Active Directory Integration, IM 36K10J40). |
+| Multi-node (2.8) | BUS | **DURM** protocol between nodes: dynamic routing, best-line selection, auto-connect to the active HAC server, encryption and server certificates. |
+| External interface (2.9) | ACCESS | ODBC (Simba server on the Core; optional license; the only route for non-Windows clients), OPC classic DA/AE/HDA (local network), OPC UA, remote database copy (mirror to external tables for complex queries; watch export size), Excel Add-in, command-line tools. |
+
+**Class, object, symbol (2.4)**
+- A **class** defines items, template items and processing; written in the *traditional*
+  language or Java. Traditional has no loops; Java classes cannot access datasets (no DSS).
+- Class structure: `PROLOG` (local variables `VAR`, parameters `ATTRIBUTE`/`PROMPT`, items
+  `SIGNAL ... FROM <template> MAKE ...`) and `BODY` (`METHOD`s with `ENTRYCONDITION`s).
+- An **object** is created from a class with parameter values; its items are created or removed
+  with it. Objects are triggered by the scheduler, trigger groups (intervals) or changes of
+  trigger signals. Traditional COMPUTE/FAST classes can also read/write datasets.
+- A **symbol** can be linked to an object so item attributes drive colour, blinking, etc.
+  Symbols with no dynamic attributes need no object.
+- Details: Class Description Language (IM 36K10P10 traditional, IM 36K10P11 Java).
+
+## Command-line tools (2.3.1, 2.9.2)
+
+| Tool | Purpose | Key usage |
+|---|---|---|
+| `dssqld` | Export / import / update / delete dataset contents; dump data dictionary | `-e <out> [<qli filter>] -d <dataset>` export (`-c` comma separated, `-a` include hidden fields); `-i <qli>` insert; `-m <qli>` update; `-r <qli>` delete; `-f` dump dictionary; `-l [<listing>]` log result; `-s <setup>` (default `dssqld.sup`). |
+| `itmval` | Read/write current item data | `-g -n <item> [-a <attr>]` get (attr 1 deadband, 2 low, 3 high, 4 lowlow, 5 highhigh); `-p -n <item> [-v <value>] [-s <status>] [-q <quality>] [-o overwrite blocked] [-t [<date>] <time> [GMT\|LST]]` set; `-e` write to event history only (host node); `-f <file> <opts>` save a command file; `-l` use local node. |
+| `ithloahis` | Load history from CSV into event-item / event-event storage groups | `-f <file> [-s <section prefix>] [-t GMT\|LST\|LCT] [-i update current value too] [-v] [-l]`. File is either `SCAN,<item1>,<item2>,...` then `<time>,<v1>,<v2>,...` (max 100 items), or `EVENT` then `<time>,<item>,<value>`. Times as date text or seconds since 1970. |
+| `ithimport` | Import item history from another CI Server | `-l` list groups; `-s -g <from> [<to>]` show; `-i -g <from> [<to>] [-f <from time>] [-t <to time>] [-c <rename file>]` import. Rename file lines: `<old name>, <new name>`. |
+| `alhimport` / `adtimport` | Import alarm / audit history from another CI Server | `-s` show, `-i` import, optional `-f` / `-t` time range. |
+
+## 3. System design
+
+**Standards to consider (3.1)**: ANSI/ISA-101 (HMI), EEMUA 191 and ISA 18.2 (alarms),
+ISA-99 / IEC 62443 and ISO/IEC 17799 (security).
+
+**Configurations (3.2)**
+
+| Layout | When to use | Watch out for |
+|---|---|---|
+| **Simple**: one CI Core does everything | Local RTUs, a few tens of clients at most | Above ~10 clients, add a separate CI Portal |
+| **Separate CI Portal** (Portal ↔ Core via DURM) | Many clients / load balancing; clients on untrusted networks; redundant Cores | Deploy display files to the Portals |
+| **HAC** (high availability): ACTIVE + STANDBY Cores | Continuous operation required; large history that needs a live copy | Stable (ideally redundant) link between servers and to the island-detection network device; license Audit log if it is monitored |
+| **Host to Host**: a client Core collects from server Cores via DURM | Reuse data at another site/level; RGS on Vnet/IP forwarding upward | Storage on the client Core can grow large |
+| **Enterprise operation/engineering** | Share data and displays across Cores; standardize engineering (EES deploys configurations) | Needs deep knowledge of dataset dependencies |
+
+- HAC: modes MANUAL / AUTO; functions monitoring (restart processes/server), watchdog (shuts
+  itself down if isolated), mirroring; options journaling (recommended on small systems) and
+  debug logging. Portals and OPC clients (via OPC tunneller) follow the active server. Dual
+  redundancy is standard; triple/quad need a special order (Tokuchu).
+- Redundant networks: DNP3, Modbus, STARDOM FCX, Host-Host (DURM).
+- Buffering and catch-up fill history gaps after outages: STARDOM FCX, DNP3, Host-Host.
+
+**Sizing (3.4)**: estimate history disk size (storage group types and lifetimes) with the sizing
+tool, and network load with the bandwidth sheets; include Vnet/IP load if connected.
+
+**Long distance (3.5)**: advise on the WAN, but the network provider (or the client's own fibre)
+owns its design.
+
+**Time synchronization (3.6)**
+- Why: Kerberos fails beyond 5 min skew; protocol timestamps; historians rejecting future data;
+  correlating events; cross-system scheduling.
+- Terms: UTC/GMT; LST (local standard, no DST); LCT (local civil, with DST). CI Server tools
+  default to LCT unless noted.
+- Adjust gently (slew, not step backwards). Forward jumps make aggregates too small, backward
+  jumps too large.
+- Topologies: simple, cascade, star, tree (segment large/wide systems).
+- Mechanisms:
+  - **(S)NTP / Windows Time**: domain members sync to the domain controller automatically;
+    workgroup machines need configuring. Registry
+    `HKLM\SYSTEM\CurrentControlSet\Services\W32Time\Parameters\Type` = `NT5DS` (domain),
+    `NTP` (use `NtpServer`, DNS name), `NOSYNC`. Poll interval via
+    `Config\MinPollInterval` / `MaxPollInterval` (2^n seconds), or
+    `NtpClient\SpecialPollInterval` when `NtpServer` ends with `,0x1`.
+  - **DURECS**: external clock on a Core serial port (Wharton 482, Trimble GPS, Arbiter
+    1092/1093), set in `durecs.sup`.
+  - **DURUTL**: distributes the time master Core's clock to other nodes over DURM
+    (`durutl.sup`: interval, three deviation limits, two adjust steps, slave node list).
+    Resolution ~2 s; cannot sync a HAC standby.
+  - **DNP3**: CI Server as master writes time to outstations (STARDOM: set `TIM_SYNC_REQ` on
+    `SD_CDNP_*_OPEN`; CI side: station *Options* › time synchronization interval).
+  - **STARDOM FCX**: CI Server writes `FT_STime` at the station's interval (needs the FCX extension).
+  - **Vnet/IP**: the CENTUM domain master distributes network time; add an SNTP server to the
+    time group on BUS-1. On Vnet/IP stations Windows Time must be stopped/`NOSYNC`.
+  - Others with time sync: IEC 60870-5-101/102/103/104, FA-M3, MELSEC, Omron FINS, Allen
+    Bradley CIP, Bristol Babcock, Fisher ROC, TIE 870-5; ProSafe-PLC via start-up command file.
+- Rules: one time master per machine; Vnet/IP overrides everything else; sync everything that
+  shares timestamped data; keep masters close; aim for under a few seconds deviation; plan
+  for time master failure.
+- History event buffering: keep *Gap time* (default 2 days) and *Maximum sample delay*
+  (default 10 days) in `itheix.sup` / `itheex.sup` well above any expected time correction.
+- Checks: `w32tm /query /configuration`, `w32tm /query /status`, `w32tm /resync /rediscover`;
+  `net time` (SMB, not NTP); `tmtime` on Vnet/IP stations.
+
+**Hardening (3.7)**
+- Windows Firewall on, with the exceptions CI Server needs; remove unused applications and
+  services.
+- Install on a fixed, tested patch level; no automatic Windows updates; apply only
+  Yokogawa-evaluated patches.
+- Antivirus with real-time scanning; exclude `C:\Users\Public\Yokogawa\tls\dat`, `…\tls\sav`,
+  `…\tls\his`.
+- Recommended user groups: **Administrators** (everything), **Maintenance** (users, logs,
+  datasets), **Engineers** (engineering module: alarms, reports, graphics), **Operators**
+  (operator screens), **InternalProcess** (internal accounts, non-expiring passwords). The
+  manual's table maps each authorization (e.g. `ITEM_MODIFY_VALUE`, `REPORT_GENERATE_ADHOC`,
+  `USER_INSERT`) to these groups; in short, read rights go to all, engineering changes to
+  Administrators and Engineers, user management to Administrators and Maintenance.
+
+## 4. Operation and maintenance
+
+**Backup and recovery (4.1.1)**
+- What to back up: the OS (~10 GB), CI Server software and setup files `%TLS_ROOT_PATH%\tls`
+  (~2 GB), database and display files `%TLS_ALLUSERS_PATH%` (~700 MB without history), and history.
+- Plan for RTO/RPO, off-site copies, labelled and rotated media, and backup result reporting.
+  Recommended: disk first tier, tape second tier; centralized backup server; never back up over
+  the control network.
+- Suggested schedule: full offline backup at version upgrades and periodic maintenance; full
+  after OS patches; weekly full; daily full or differential of the `tls` folder; archive history
+  weekly; export key datasets daily with `dssqld`.
+- Don't change configuration during a backup. Use VSS for consistent online copies.
+- Consistency tools: `ismchk <file>` / `ismkfr [-c] [-r] [-s] [-d] <file>` for ISAM files;
+  `dsschk [-d <dataset>] [-n <node>] [-e] [-r] [-v]` for replicated/duplicated datasets.
+  Repair: repeat `dsschk -r` until clean, then `dsschk -e`, then `dsschk -e -r` (maybe twice).
+- Disaster recovery: restore system backup (or reinstall OS + CI Server + patches) → restore
+  `tls` configuration and database → verify.
+- Retrieve archived history from the Engineering module (you can set its lifetime).
+- Restore lost records: back up current data, build a `.qli` from the daily `dssqld` export,
+  import.
+
+**Routine (4.1.2–4.1.5)**: OS patches, antivirus updates, user account housekeeping; watch
+Windows and CI Server Performance Monitor, Windows event log, CI Server system log report and
+audit trail report.
+
+**Monitoring processes (4.1.6)** — INTEGRATION Service tools, each configured in its `.sup` file
+and reporting status into items:
+
+| Process | Monitors | How to start |
+|---|---|---|
+| `intmon` | process running; DURM node and channel status (can set slave items OFFLINE) | Uncomment its line in `%TLS_ROOT_PATH%\tls\com\integration_start.cmd` |
+| `intcpumon` | CPU load (any perf counter) | `start /B /NORMAL %TLS_ROOT_PATH%\tls\exe\intcpumon -n intcpumon` then `durwfp -p intcpumon -t 10` |
+| `intdskmon` | disk free space | same pattern with `intdskmon` |
+| `inttcpmon` | TCP reachability (ping) of hosts | same pattern with `inttcpmon` |
+| `intadtnot` | writes an AUDIT event from a script | `intadtnot -e A.B.C.D [-u user] [-t terminal] [-o old] [-v new]` |
+| `intspawn` | runs a command on item value/status/option change, timer or node up/down (max 64 triggers) | `intspawn -n intspawn`, triggers in `intspawn.sup` |
+
+**Problem handling (4.2)**: identify → investigate (performance monitors, logs, debug mode,
+debug tools) → fix (memory size, queue size, load balancing, patch).
+
+## 5. Implementation
+
+**Engineering tools (5.1)**: License Request / Authorization wizards, Setup File Editor
+(changes apply after the process restarts), Engineering module (Enterprise version for
+enterprise systems), Edit module (displays and symbols), CI Exchange (import CENTUM VP
+projects), Performance Monitor, Message log, ASPA, Excel Add-in, `hacw_hmi` (HAC management).
+
+**CENTUM VP import (5.2)**: CI Exchange reads the VP project → produces import files and an
+import command → run it → review scan times against Vnet/IP load.
+
+**Designing classes and symbols (5.3)**
+1. Create template items (analogue, digital, string…) in a common system section.
+2. Group devices of the same type, define their inputs/outputs and common behavior in a class.
+3. Decide which items the class creates; creating an object fails if its items already exist.
+4. Group equipment with the same look into one symbol.
+5. Sections in a class are relative to the object's section. Template items must exist before
+   objects are created (matters for migration scripts).
+6. For performance, prioritize objects.
+
+**Testing checklist (5.4)**: input validation (valid/invalid, boundaries, combinations,
+defaults), outputs (destination, value, format, messages), interfaces (including offline
+partners), scheduled tasks, UI (initial values, invalid/quick/concurrent operations,
+usability), administrator tasks (backup, accounts, monitoring), performance and long-run
+(leaks, disk), security (privileges), localization, regulations.
+
+**Tuning (5.5.1)**
+- Common memory: max processes 250 by default (`dur.sup`); keep free common memory above 10 %
+  and usage below 75 % (item pool size in `itm.sup`).
+- Queues must not hold waiting requests; enlarge in each process's setup file. Objects running
+  too often: fix triggers/schedules, or add instances and priorities.
+- Database: collisions under 250/s; entries under 80 % of max users (default 2K); max files
+  1,200. Set `-u` / `-f` in `database_start.cmd`; check with `ismdbg`.
+
+**Debugging (5.5.2)**
+- Enable debug/log per process in its setup file (`<process>.sup`, level 1–3, log size);
+  e.g. `alnmgr`, `eqp*` (`eqpfcx.sup`, `eqpmdc.sup`, or `eqpdbg` until restart), `hac`, `opc`
+  (PRINT statements to UMHLOG, on by default in `opc.sup`), `opxdac` (`opxsys.sup`),
+  `opxuac_mon`. Turn debug off afterwards (performance).
+- Debug-mode launchers: `jwuipoi_debug` (operator interface), `jhmiset_debug` (Engineering
+  module), `jwuiged_debug` (Edit module), `fsuedit_debug` (Setup File Editor), `ftmoni_debug`
+  (Performance Monitor), `jwuitrd_debug` (Trend), `jwuiaan_debug` (ASPA),
+  `jtoolslic_debug` / `jtoolslicreq_debug` (license wizards).
+- Debuggers: `dssxdb`, `dsschk`, `ismdbg` (DATABASE); `hisdbg`, `hasdbg` (HISTORY);
+  `itmdbg`, `ithdbg`, `ithrecover` (ITEM); `opxxdb` (ACCESS); `rptgen` interactive (REPORT);
+  `opcdbg` (PROCESS); `almdbg` (ALARM); `eqpdbg` (EQUIPMENT).
+
+**Migration (5.6)**, e.g. from a demo-license system to the target system
+1. Export master datasets with `dssqld -e`, using `.qli` filter files (`@<DATASET>`,
+   `@FIELD`, `@SELECT`) where needed: `AUTH_GROUP_DF`, `USER_DF`, `PRINTER_*_DF`,
+   `STATUS_DF`, `ALARM_ACK_DF`, `ALARM_AOI_DF`, communication `*_LINE_DF` / `*_STATION_DF` /
+   `*_POINT_DF`, `INSTALL_DF`, `UNIT_DF`, `ITEM_DF`, `HIS_GROUP_DF` (item storage groups
+   only), `CLASS_DF`, `OBJECT_DF`, `REPORT_DF`. Drop fields removed in the new release.
+2. Re-apply setup file changes on the new system (setup file contents can change between releases).
+3. Edit exports (e.g. node names in `PRINTER_TYPE_DF`, `*_LINE_DF`, `ITEM_DF`; set the
+   `COMPILE` flag to 1 in `CLASS_DF`), then import in dependency order:
+   template items → classes (compiled) → objects that create items → remaining items →
+   objects that create none. Check each `.qlo` listing for errors; reorder on relation errors.
+4. History: copy `%TLS_HIS_PATH%` (except `has`) and `his_group.*` / `his_units.*` from
+   `%TLS_DAT_PATH%`; create the same item storage groups; `ithimport -g <group> -c <rename file>`;
+   `alhimport -i`. Alarm tag names cannot be renamed.
diff --git a/knowledge/progress.md b/knowledge/progress.md
new file mode 100644
index 0000000..eccb977
--- /dev/null
+++ b/knowledge/progress.md
@@ -0,0 +1,86 @@
+# Progress
+
+Status of each manual and each practiced task. Update this file as work proceeds.
+
+**Manual status:** not started · summarized · in practice · done
+**Task status:** not practiced · practiced, awaiting confirmation · passed · not completed (see notes)
+
+## Test system
+
+Practice notes refer to the repo author's **development test system**, not to anything shipped
+with this repo:
+
+- CI Server R1.03.00, unlicensed, on the development machine.
+- **AID** (e.g. `AID.WRPS.SIM.INFLOW`) is the author's own test project on that machine. It is
+  not in this repo and other systems will not have it. Item names here are evidence of what
+  was tested, not names to use elsewhere.
+- Communication stations seen there (`WRPS_PLC`, `MODSIM_STATION`, `STN_MQTT`) belong to AID.
+
+Shared, repeatable practice will use the sample project **CI-Eng-Sample**
+(`examples/CI-Eng-Sample/`) once it exists.
+
+## Manuals, in working order
+
+| Phase | Manual | Title | Status |
+|---|---|---|---|
+| 0 | IM 36K01A50 | Release Information | summarized ([notes](manuals/IM36K01A50-01EN.md)) |
+| 1 | IM 36K10J10 | System Design/Management | in practice, skill `system-management` ([notes](manuals/IM36K10J10-01EN.md)) |
+| 2 | IM 36K01C10 | CI Server Setup | not started |
+| 2 | IM 36K01C30 | Security Guide | not started |
+| 3 | IM 36K10G10 | Tutorial | not started |
+| 4 | IM 36K10R10 | CI Portal Setup | not started |
+| 4 | IM 36K10R20 | Setup File Editor | not started |
+| 5 | IM 36K10C10 | Engineering Module - I/O Processes | not started |
+| 5 | IM 36K10L40 | EQUIPMENT Service - Utilities | not started |
+| 5 | IM 36K10L41 | EQUIPMENT Service - Configuration | not started |
+| 5 | IM 36K10L35 | ITEM Service - ITH | not started |
+| 5 | IM 36K10L36 | ITEM Service - ITM | not started |
+| 6 | IM 36K10L10 | Class/Object Engineering | not started |
+| 6 | IM 36K10P10 | Class Description Language - Traditional | not started |
+| 6 | IM 36K10P11 | Class Description Language - Java | not started |
+| 7 | IM 36K10L55 | Exchange Tool Engineering | not started |
+| 7 | IM 36K10L30 | BUS Service | not started |
+| 7 | IM 36K10L45 | Data Access Engineering | not started |
+| 7 | IM 36K10E20 | OPC Interface | not started |
+| 8 | IM 36K10L15 | Alarm Management Engineering | not started |
+| 8 | IM 36K10C20 | Engineering Module - Alarms | not started |
+| 8 | IM 36K10E30 | Alarm System Performance Analysis | not started |
+| 9 | IM 36K10L20 | History Management Engineering | not started |
+| 9 | IM 36K10C40 | Engineering Module - Historical | not started |
+| 10 | IM 36K10R30 | Graphic Display Setup | not started |
+| 10 | IM 36K10C50 | Edit Module | not started |
+| 10 | IM 36K10C60 | Advanced Operating Graphics | not started |
+| 11 | IM 36K10L50 | Data Set Service Engineering | not started |
+| 11 | IM 36K10P30 | Data Set Service Language | not started |
+| 11 | IM 36K10P20 | Report Generation Language | not started |
+| 11 | IM 36K10R40 | Report Server Setup | not started |
+| 11 | IM 36K10C30 | Engineering Module - Reporting | not started |
+| 11 | IM 36K10E10 | Excel Add-in | not started |
+| 12 | IM 36K10L25 | Audit Trail Management | not started |
+| 13 | IM 36K10J20 | High Availability Computing | not started (needs two servers) |
+| 13 | IM 36K10J40 | Active Directory Integration | not started (needs a domain) |
+| — | IM 36K01A10 / A20 / A30 | Read Me First, License, OSS terms | skipped (no engineering content) |
+
+Order change after reading IM 36K10J10: its migration procedure (§5.6) shows the dependency
+chain *communication lines/stations/points → installations/units → items (template items
+first) → storage groups → classes → objects → reports*. So acquisition and items (phase 5)
+now come before classes and objects (phase 6), and the other interfaces follow as phase 7.
+
+## Tasks
+
+| Manual | Task | Status | Notes |
+|---|---|---|---|
+| J10 | System check: version, paths, running state, time sync ([skill](../.claude/skills/system-check/SKILL.md)) | passed | Started CI Service elevated; live version R1.03.00; node UNLICENCED (expected). |
+| J10 | Time sync check / configuration (`w32tm`, `durutl.sup`) | not practiced | Check only on this system; no Vnet/IP. |
+| J10 | Export a dataset with `dssqld` ([task](../.claude/skills/system-management/tasks/export-dataset.md)) | passed | Filter-file syntax: [ISS-003](issues.md). |
+| J10 | Import / update a dataset with `dssqld` | not practiced | Writes to the database; ask first. |
+| J10 | Read an item with `itmval -g` | passed | Read `AID.WRPS.*` items. `-a 1..5` gave `ITM-E-WRONG_ATT` on an item without limits: [ISS-004](issues.md). |
+| J10 | Write an item with `itmval -p` | practiced, awaiting confirmation | `AID.WRPS.SIM.INFLOW` 419.76 → 400 → 419.76. After the first write the item's options gained the `uoff` flag (0x8a → 0x8e), which stays after restoring the value. See [ISS-001](issues.md). |
+| J10 | Read the message log (`tls\log\umhlog.log`) | passed | |
+| J10 | Per-component version list (`tls\lst\tls_versions.lst`) | passed | |
+| J10 | Load history from CSV with `ithloahis` | not practiced | Needs an item storage group. |
+| J10 | Consistency check (`dsschk`, `ismchk`) | not completed | `dsschk -d ITEM_DF -v` exits 0 with no output and nothing new in the message log; See [ISS-002](issues.md). |
+| J10 | Monitoring processes (`intmon`, `intcpumon`, `intdskmon`, `inttcpmon`, `intspawn`) | not practiced | Needs items to report into; better after phase 5. |
+| J10 | Backup of `tls` folders | not practiced | Plan only; the actual backup tool is site-specific. |
+| J10 | Migration between systems (§5.6) | not practiced | Needs engineered data; revisit after phase 6. |
+| J10 | Tuning and debugging (setup files, `*_debug` launchers, debuggers) | not practiced | |