Everything in the atlas

It writes and it administers — ETL extraction and database user management are not read operations, and the same connection that profiles a column can move data out of the tenant. Masking is only as good as the configuration naming the fields, so an unlisted sensitive column is an unmasked one.

Its README says it plainly and so does this entry: this is not a simulation, every tool call reaches a live BW system, and write operations produce real changes. BW modelling objects are the definition of a warehouse — the blast radius of a wrong one is every report built on it.

It reaches a data warehouse, so what an agent can see is business data by definition — the SQL sanitisation and PII filtering in the server are there because of that, not instead of a decision about it. And the guidance is Datasphere-shaped: nothing here transfers to a HANA Cloud database used directly.

Ninety tools is a large surface to put in front of a model at once, and client-credentials OAuth means the tenant sees one identity rather than the person asking. Content transport and user management sit in the same tool list as reading a story.

When user-dependent responses are cached without user-aware or explicit operation keys, or translated responses share keys across locales. These cases require key isolation; locale-aware keys are off by default.

It writes and it administers — ETL extraction and database user management are not read operations, and the same connection that profiles a column can move data out of the tenant. Masking is only as good as the configuration naming the fields, so an unlisted sensitive column is an unmasked one.

It reaches a data warehouse, so what an agent can see is business data by definition — the SQL sanitisation and PII filtering in the server are there because of that, not instead of a decision about it. And the guidance is Datasphere-shaped: nothing here transfers to a HANA Cloud database used directly.

Where the release and programming model offer durable locks, as RAP draft handling does, the README states that those solve the problem at the kernel level and the lock manager is for where that is not an option. A running handler occupies one background work process for as long as any lock is held, and only generated `ENQUEUE_*` modules can be locked.

Not an end-user app: it reads from any table the user names and writes into tables of the Z/Y namespace, with no authorisation check of its own, so the README asks a project to add its own checks on `S_TABU_DIS`/`S_TABU_NAM` and to restrict who may run it before it is used beyond a development system. XLSX upload and download are not ready for ABAP Cloud yet, and of the upload apps only the JSON one writes the file's rows back to the database.

It writes and it administers — ETL extraction and database user management are not read operations, and the same connection that profiles a column can move data out of the tenant. Masking is only as good as the configuration naming the fields, so an unlisted sensitive column is an unmasked one.

Its README says it plainly and so does this entry: this is not a simulation, every tool call reaches a live BW system, and write operations produce real changes. BW modelling objects are the definition of a warehouse — the blast radius of a wrong one is every report built on it.

Ninety tools is a large surface to put in front of a model at once, and client-credentials OAuth means the tenant sees one identity rather than the person asking. Content transport and user management sit in the same tool list as reading a story.

Nothing about it is readable from outside the package: there is no repository, no issue tracker, and the npm page carries no readme, so the README, the examples and the changelog can be read only from the tarball or a CDN mirror, and a bug has nowhere to be filed. Provider URLs are called from the plugin with API keys taken from environment variables in a .env file, so on BTP no call resolves through a destination and the keys live in the environment. The npm line stops at 1.0.1 on 3 November 2025; the changelog shipped inside that version is headed v2.1, so the version npm shows and the version the author counts are not the same.

Not for a table whose contents matter, and not for one chosen at run time: the name of the table is fixed in the calling code, and the maintainer's own warning is "only hardcode the table name, don't use with important tables, use at own risk!"

The app writes to whatever table is passed in the URL parameter `table` and carries no authorisation check of its own, so past a development system it needs authorisation checks added and a restriction on who may run it.

The metadata is inconsistent: the declared repository is github.com/acme/search-vocab and the README installs @acme/search-vocab, neither of which matches the published name. One release, November 2025.

It changes what DELETE means for every client of the annotated projection, so anything that must actually remove a row — an erasure request, a retention job — has to go around it.

Any run nobody is watching: the drawing tools work only while the draw.io canvas at `http://localhost:3000/` is open and connected, so there is no diagram to produce on a build agent or a server. Nothing checks the result either, because the maintainer keeps a diagram linter on the [roadmap](https://github.com/etosin/arc-draw#-roadmap) rather than in the release. The maintainer marks the status early at v0.x, installs it from source, and records testing with Claude Code alone.

The README tells you to fork before cloning, because Claude rewrites `references/` and `SKILL.md` in place as you use it, so a clone shared between people changes under all of them. Installing means `git` and `ln` on macOS or Linux and a symlink into `~/.claude/skills`. Fine positioning stays manual: the coordinates Claude writes can be rough apart from the centre coordinates of elements joined by a straight arrow and the containment relationships.

A diagram of anything other than an SAP, BTP or on-premise landscape: the skill is tuned to the Horizon style and its 71 templates. It drafts rather than finishes, and its README budgets 15 to 30 minutes of edits in draw.io per diagram. Without Python the validator does not run.

A system below ABAP 7.50: some of the classes it uses are only available from 7.50, and the maintainer states that downward compatibility to 7.40 in a single branch is currently not possible without resorting to dynamic programming.

Not on ABAP Cloud: the system tables and the classic Workbench APIs the screens use are not released for the ABAP Cloud language version, and there is no 7.02 version below the 7.50 floor either. Not where the full original function is needed, since only the listed transactions exist and about 300 buttons and fields are shown disabled. The authorization checks are rebuilt rather than the originals, which is why the README asks that the screens be tried with a restricted user on a sandbox before the ICF service is opened to more people.

Not as it stands on a system past development: it reads the contents of any table the user names without an authorisation check of its own, so access is bounded only by whatever restrictions exist on the underlying handler or service. The README asks for your own checks first, naming `AUTHORITY-CHECK` on `S_TABU_DIS`/`S_TABU_NAM`, and for a restriction on who may run the app. There is no 7.02 downport either, because the selection-screen dependency has no `702` branch.

It runs whatever SQL a user types without an authorisation check of its own, and the native path additionally uses ADBC, which bypasses ABAP authorisations and client separation; the README asks anyone running it beyond a development system to add authorisation checks and restrict who may start the app. The native console needs Standard ABAP, so on ABAP Cloud only the Open SQL console is available.

It targets the Application Frontend service only — not the HTML5 application repository — fixes authentication to IAS, and always creates new service instances, so reusing existing ones means editing the generated `mta.yaml` afterwards. Version 0.1.0 is the only release.

The version list comes from ui5.sap.com, so a build that cannot reach it has nothing to match the project's version against. It checks the version written in `ui5.yaml`; a framework version a deployed app loads from elsewhere is outside what it sees.

It audits through npm: a project that resolves its dependencies with another package manager, or has no `package.json`, has nothing for it to run. What it reports is what `npm audit` knows, so a vulnerability outside the npm advisory data is not in its table.

A `SELECT` from a business table has nothing to select from: the database in the page holds the framework's own tables, and nothing can reach outside the browser, so an app that reads a system's data cannot be tried here. Only the UI5 libraries built in are available and a control from anywhere else will not load; what the transpiler cannot compile is reported in the output panel rather than run. State lives in the tab — a reload, and a press of Run, starts over on a fresh database.

Not below SAP_BASIS 7.40 SP08, and not on [SAP S/4HANA Cloud Public Edition](external:sap-s4hana-cloud-public-edition) or [the SAP BTP ABAP environment](artifact:btp-abap-environment), where the classic transport check, the report and job calls and the cluster table `INDX` it rests on do not exist and the maintainer names gCTS as the answer instead. Only on a development system where every developer is trusted: each developer's GitHub token is kept in a cluster table encrypted with a key held in the same table, so anybody who can run ABAP on that system can read both, and the separation between author and reviewer holds only as far as that trust does. Not where the gate has to be unconditional — `#NOREVIEW` in a request's short text skips the check for that request, `ZTR_DISABLED` switches it off for a system, a check that cannot reach GitHub lets the release go on, and nothing checks who writes the marker.

A system that reports itself productive, or whose role the probe cannot prove non-productive, is write-locked with no override, so changing code in production is not something this does. The permission ladder constrains the package, object name and transport request the server itself passes on a write, and is not a sandbox around ABAP it executes: `abap_run`, `abap_test` and the opt-in `core.eval` run under the connected technical user's SAP authorisations, which is where the boundary actually sits. Data preview, dump variable contents and driving a transaction by batch input stay off in every mode, `admin` included, until each is named explicitly.

Driving the GUI reaches whatever a user can reach, which is business data and not a development surface. Decide what an agent may do in a session before granting one, not after. Windows only. The HTTP transport is meant for team use but binds to localhost, and unattended multi-user operation is listed as not yet available — a limitation and also the safer default.

Driving the GUI reaches whatever a user can reach, which is business data and not a development surface. Decide what an agent may do in a session before granting one, not after. It has also not been touched since February 2025, which the maintenance reading states — the three later entries here are where the CI, the Web GUI and the testing framing went.

Not a tool a developer runs: it is a library to build one with, and it ships no editor, command line or interface of its own. Its sample connects with a plain HTTP URL, a user and a password, so a system that does not expose the adt ICF service that way is out of reach.

The embedded preview needs a system that accepts basic auth: the documentation states that pure SSO or SAML without a basic-auth fallback is not supported, and points at `external` mode, which opens the app in the normal browser with the existing SAP session. On the web only the language half runs; the preview with its proxy, traffic log and screenshot, Ctrl+F3 activation, the render gate, the workspace-wide check and the MCP servers are desktop-only. The agent app tools send nothing until `abap2ui5.agent.enableAppTools` is switched on in User settings, because an event they fire runs as the configured user and changes data on the system for real.

A `SELECT` from a business table has nothing to select from: the database in the page holds the framework's own tables, and nothing can reach outside the browser, so an app that reads a system's data cannot be tried here. Only the UI5 libraries built in are available and a control from anywhere else will not load; what the transpiler cannot compile is reported in the output panel rather than run. State lives in the tab — a reload, and a press of Run, starts over on a fresh database.

It declares a VS Code engine and binds to the .tabl.ddic file extension, so it does nothing for a table definition opened in another editor, or read from a system rather than from a file on disk.

Not on ABAP Cloud: the system tables and the classic Workbench APIs the screens use are not released for the ABAP Cloud language version, and there is no 7.02 version below the 7.50 floor either. Not where the full original function is needed, since only the listed transactions exist and about 300 buttons and fields are shown disabled. The authorization checks are rebuilt rather than the originals, which is why the README asks that the screens be tried with a restricted user on a sandbox before the ICF service is opened to more people.

The class has to be moved by hand: an abapGit pull of the exported ZIP, or a paste into ADT or SE24. The palette is held to the 121 controls of UI5 1.71, and a control inserted by its class name instead comes from whatever UI5 the canvas has loaded, with nothing checking it against 1.71 - *Checks* reports that. An app class that builds no view, or more than one, is refused.

The exported file list is hard-coded, so a dev space whose setup lives in files the readme does not name is not carried across; the roadmap lists making the list flexible as open work. Nothing from the BAS profile is in the zip, so a developer who wants the colour scheme, the keymaps or the installed extensions uses the BAS profile export on its own.

Windows only, with PowerShell 7 (`pwsh`) installed; the run needs access to `download.eclipse.org`, `tools.hana.ondemand.com`, `eclipse.devepos.com` and the update site of every plugin selected, from the machine it installs on. A plugin outside the catalog is installed only after someone adds its update site and feature group id to `catalog.json`.

Everything it checks is a name: a file name under `rootDir`, or an identifier a parser extracted from a file, so a convention that cannot be written as a regex over one name has no rule shape here. Content rules targeting `.hdbanalyticalprivilege` or `.hdbconstraint` return no results, which the README states as not yet implemented, so conventions centred on those two artifact types go unchecked.

There is no repository and no README: what the eight commands do is only readable from the compiled dist that ships in the package. Three of them — npmrc, install, project — are written for "SimpleMDG" projects rather than for CAP generally, so the reusable part is bind-local, run-local and debug. Seven releases up to July 2026. Its binding step writes the resolved VCAP_SERVICES into default-env.json in the project, where [@sap/cds-dk](artifact:sap-cds-dk) keeps only the place the credentials can be fetched from, so a project using it needs that file kept out of version control.

A `SELECT` from a business table has nothing to select from: the database in the page holds the framework's own tables, and nothing can reach outside the browser, so an app that reads a system's data cannot be tried here. Only the UI5 libraries built in are available and a control from anywhere else will not load; what the transpiler cannot compile is reported in the output panel rather than run. State lives in the tab — a reload, and a press of Run, starts over on a fresh database.

It is not the framework and takes no framework fixes: those belong upstream in abap2UI5, and nothing under `src/`, `downport/`, `output/` and `build/` here is a source file. The build is frozen on pinned commits with no scheduled rebuild, so it does not show what abap2UI5 main does today, samples in subfolders of the samples repository are excluded because they require a real SAP system, and a known upstream transpiler issue leaves `check_on_init( )` always false, so backend-built view XML may not be returned.

On a system where the abap-search-tools backend is not installed: the plugin lists it as a precondition, and the beta update site expects that backend's dev branch in particular.

On a system below ABAP 7.51, or one without a connected HANA database, the client-based mode does not run, and the server-based mode is only available where the abap-code-search-tools backend is installed in the system.

On a system where the abap-search-tools backend is not installed: the plugin lists it as a precondition, and the beta update site expects that backend's dev branch in particular.

Not on ABAP Cloud: the system tables and the classic Workbench APIs the screens use are not released for the ABAP Cloud language version, and there is no 7.02 version below the 7.50 floor either. Not where the full original function is needed, since only the listed transactions exist and about 300 buttons and fields are shown disabled. The authorization checks are rebuilt rather than the originals, which is why the README asks that the screens be tried with a restricted user on a sandbox before the ICF service is opened to more people.

A Python 3.10+ CLI, installed with pipx rather than through any SAP-provided package manager. Its own README lists what it does not do yet: where-used, syntax check and unit test runs — for those it says to reach for [ADT/Eclipse](external:eclipse) or an MCP server. First release, v0.1.0, and no standalone LICENSE file in the repository — the MIT grant is in the README and in pyproject.toml's project metadata rather than its own file.

Not below SAP_BASIS 7.40 SP08, and not on [SAP S/4HANA Cloud Public Edition](external:sap-s4hana-cloud-public-edition) or [the SAP BTP ABAP environment](artifact:btp-abap-environment), where the classic transport check, the report and job calls and the cluster table `INDX` it rests on do not exist and the maintainer names gCTS as the answer instead. Only on a development system where every developer is trusted: each developer's GitHub token is kept in a cluster table encrypted with a key held in the same table, so anybody who can run ABAP on that system can read both, and the separation between author and reviewer holds only as far as that trust does. Not where the gate has to be unconditional — `#NOREVIEW` in a request's short text skips the check for that request, `ZTR_DISABLED` switches it off for a system, a check that cannot reach GitHub lets the release go on, and nothing checks who writes the marker.

Where abapGit itself is not installed in the system, since every tool here acts on its repositories.

Event roundtrips and visual fidelity stay with a live run: the documentation leaves those to the project's MCP server and to [abap2UI5 for VS Code](artifact:abap2ui5-vscode-extension). A class that builds view parts in helper methods without the handle idiom cannot be statically reconstructed, and the render gate is then skipped with a notice while the property gate runs on what was reconstructed. A screenshot is not a preview of the app: nothing round-trips, no event reaches ABAP, and a table filled by a SELECT photographs empty unless a mock JSON file sits beside the source.