[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C2M9XVTRVXYXMGTEC62EFB":3,"skill-same:01M7C2M9XVTRVXYXMGTEC62EFB":53},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":6,"ownerHtmlUrl":27,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":28,"install":29,"resources":42},"01M7C2M9XVTRVXYXMGTEC62EFB","security",null,"Use when deciding who may read what in Denodo 9.5 — masking or hiding columns (PII, contact details, salaries, bank accounts) from some users or roles, letting a role see only some rows (a country, a tenant) or each person only their own, taking someone's access away (ALTER USER … REVOKE), denying a view to a group, a global security policy (CREATE GLOBAL_SECURITY_POLICY) and the VDP tags it reads, giving a user a role or giving a role access to a database or views (CREATE ROLE … GRANT EXECUTE, ALTER USER … GRANT ROLE), checking what a given user really gets back, or a user who sees data a restriction should hide (\"he still gets every row\", \"the masking stopped working\"). Not for a tag as a label without a policy (\u002Fdenodo:catalog) or for descriptions and the MCP visibility tag (\u002Fdenodo:semantics).","kirivandubai\u002Fdenodo_wipe_coading_skills","kirivandubai","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F221521416?v=4","01M7C2HC9KETA166XKR09YMKZQ","01M7C2HCAJ787NNYKMCG12S3DY",0,"MIT","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fsecurity\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: security\ndescription: Use when deciding who may read what in Denodo 9.5 — masking or hiding columns (PII, contact details, salaries, bank accounts) from some users or roles, letting a role see only some rows (a country, a tenant) or each person only their own, taking someone's access away (ALTER USER … REVOKE), denying a view to a group, a global security policy (CREATE GLOBAL_SECURITY_POLICY) and the VDP tags it reads, giving a user a role or giving a role access to a database or views (CREATE ROLE … GRANT EXECUTE, ALTER USER … GRANT ROLE), checking what a given user really gets back, or a user who sees data a restriction should hide (\"he still gets every row\", \"the masking stopped working\"). Not for a tag as a label without a policy (\u002Fdenodo:catalog) or for descriptions and the MCP visibility tag (\u002Fdenodo:semantics).\n---\n\n# Who may read what\n\nEverything here is server-wide, and it changes what real people get back from views —\nsilently. The statement succeeds; then an analyst's report shows `********` where a name\nwas, an amount comes back `NULL`, one analyst still reads every row, or a policy stops\nbeing evaluated without anyone being told. Two facts shape every step below:\n\n1. **Administrators are never restricted** — global administrators, and local administrators of\n   the view's database (`ADMIN` on it). The profile you connect with may be one: `env check` →\n   `vdp.admin`. **Your own `SELECT` proves nothing about a restriction.**\n2. **The proof is the same query run as the person** — `CONTEXT ('impersonate_user' = '\u003Cuser>')`,\n   or `('impersonate_roles' = '\u003Crole>')` — with their privileges and every policy, and no password.\n   It needs `vdp.impersonation: true` (the profile's user has the `impersonator` role).\n\nA tag as a plain label is `\u002Fdenodo:catalog` (its full syntax lives there); descriptions and the\nMCP visibility tag are `\u002Fdenodo:semantics`. User accounts and passwords, LDAP and\nidentity-provider groups, per-role row and column restrictions, custom policies,\nsession-attribute audiences and server properties are set in Design Studio by the\nadministrator — say so and stop. Applying files is `\u002Fdenodo:execute`; the working loop and the\ncore safety rule are `\u002Fdenodo:vql`.\n\n## Before you change anything\n\nReads only, and every decision below comes from them — complete only for a global administrator\n(`vdp.admin: true`). For anyone else, a local administrator too, query 1 silently returns only the\nprofile's own grants and its roles': say the list of readers is partial.\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\n-- 1. Who reads the database, and through what: one row per grant. A user's row with\n--    userrolename empty = granted to that user directly; a row with username empty = a\n--    role's own grant (a role nobody holds shows only this way); elementname empty = the\n--    whole database\nSELECT username, userrolename, elementname, dbconnect, dbexecute, elementexecute, dbadmin\nFROM CATALOG_PERMISSIONS() WHERE dbname = 'sales_analytics';\n\n-- 2. Policies already on the server, and whether each is still evaluated (INVALID = not)\nSELECT name, element_status, description FROM GET_ELEMENTS() WHERE type = 'globalSecurityPolicy';\n\n-- 3. Which policy reads which tag\nSELECT global_security_policy_name, tag_name FROM GET_GLOBAL_SECURITY_POLICIES_TAGS();\n\n-- 4. Tags already on the views of the database\nSELECT view_name, column_name, tag_name FROM GET_VIEW_TAGS() WHERE input_database_name = 'sales_analytics';\n```\n\n- Then `DESC VQL GLOBAL_SECURITY_POLICY \u003Cname>` for each policy that names a tag, a role or a\n  database you are about to touch — its audience is not in any of the queries above. Like query 3,\n  it needs an administrator of the database. It starts with `DROP … CASCADE`: read it, never apply it.\n- **A tag that a policy names is a security switch**, not a label: adding it to a column\n  restricts that column for the policy's audience, removing it lifts the restriction.\n- One user's grants in the database: query 1 with `AND username = '\u003Cuser>'`. Not `DESC VQL USER`:\n  without `('includeUserPrivileges' = 'yes')` it shows no roles at all, and with it, for a\n  local user, it prints the password hash into the transcript.\n- Someone known only to LDAP, Kerberos or an identity provider is no VDP user: no row in query 1, and\n  `impersonate_user` says `does not exist`. Their access is the roles of their groups: ask the human\n  which, check them with `impersonate_roles`, and hand the check of a per-person filter to the human.\n- Whether a role exists: `DESC ROLE \u003Cname>` (`Error loading role` when it does not).\n  `GET_USERS_WITH_ROLE()` answers a missing role with 0 rows and no error, and `LIST ROLES`\n  is cut at the tool's 100 rows (`truncated: true`) unless you pass `--max-rows 5000`.\n\n**A user sees what they should not** — in this order, each step a read:\n\n1. Query 2: is the policy `INVALID`? `DESC VQL` of it ends with `# Invalid object policy`;\n   then `DESC ROLE` each role of its audience — one that is gone switched the policy off for\n   all of them (Silent failures, 6).\n2. Query 1 for that user: a row with `userrolename` empty, a role outside the audience, or\n   `dbadmin = true` is a path the policy does not restrict (Choosing the audience).\n3. Query 4: is the tag on the lowest view the user can read, and on the column the view\n   exposes? `SELECT column_name, dependency_name, depth FROM COLUMN_DEPENDENCIES() WHERE\n   input_view_database_name = '\u003Cdb>' AND input_view_name = '\u003Cview>'` names the views and the\n   base column behind each output column (`\u002Fdenodo:views`, lineage) — *verified: 9.5.1 (live, 2026-10-05)*.\n4. The impersonated query as that user, now and after each fix you propose.\n\n## Who applies what\n\n| What the statements touch | You apply yourself | Only after the human's yes |\n|---|---|---|\n| only objects you created in this session: a role no person holds yet, whose grants name only objects you created in this session, a policy limited by `VIEW_DATABASES` to a database you created in this session and whose audience is such a role, tags on views you created | create, change, check | — |\n| anything that existed before this session — a role, a user, a tag a policy names, a policy, someone else's view; any grant or revoke of a role or a privilege to a person; a new policy whose audience or views include existing ones; re-creating a role someone dropped | the reads, the impersonated checks, the files | every statement |\n\n`vql plan` on the file names, per statement, every existing object it touches that this\nsession did not create (`touches`) — the first row needs that list empty (`\u002Fdenodo:vql`).\nThe tool marks `CREATE` of a role, a user or a policy `destructive: security`.\n\nThe yes is to the statements, after you have shown them in this shape:\n\n```\nDatabase sales_analytics. Readers today (CATALOG_PERMISSIONS): role sales_analyst — mlee,\nrpatel; kchen also has EXECUTE on household_income granted to her directly.\n1. security\u002Fpersonal_data.vql: tag personal_data, policy analyst_masking (role sales_analyst)\n2. model\u002Fiv_household_income.vql: the view's file gains TAGS on buy_potential, dependents\nAfter: mlee, rpatel see buy_potential as ********, dependents as NULL — in every view on\niv_household_income; avg_dependents in household_income_by_band becomes NULL for them.\nNot covered: kchen still reads both in clear (direct grant). Revoking it is a separate yes.\nAlso exposed, not asked about: the role reads every column of the base views below.\nApply 1 and 2?\n```\n\nWhen you cannot ask — the human is away, the deadline is close — the answer is the files and\nthis message, with the impersonated checks the human can run after applying, not the\nstatements applied. Say what waiting costs, and which file goes first. A file waiting for a\nyes keeps the template's `verified:` mark — the template was run, this file was not — and\nthe message says it is not applied. It waits whole: a statement the plan calls yours (a new tag)\nwaits with the rest, or it is left behind with nothing to do. Run its impersonated checks now anyway: today's answers\nare the \"before\" the human compares with.\n\n| Rationalization | Reality |\n|---|---|\n| \"They named the exact columns and the treatment — that is the yes\" \u002F \"they said *set it up and make sure it works*\" | They named an outcome. They have not seen that a number becomes `NULL` in every report, nor which user the policy leaves out. The yes comes after the statements. |\n| \"Every object is new, carries my prefix and is limited to one database\" | A name you chose does not make what it touches yours. The policy restricts the people who read that database, the tag sits on someone else's view: the first row of the table needs the database and the views created by you in this session, not the policy. |\n| \"It only restricts — it can only make things safer\" | A mask changes what reports compute: a `GROUP BY` on the masked column collapses into one group, an aggregate over a hidden number becomes `NULL`, a filter on it finds nothing. Someone's dashboard breaks silently in the morning. |\n| \"It is reversed with one `REMOVE_FROM` \u002F `DROP`\" | The readers already got the result; and `DROP TAG … CASCADE` deletes every policy that names the tag. |\n| \"Only new objects — a pure `CREATE`\" | A new policy restricts existing people; a re-created role revives an invalid policy; `CREATE OR REPLACE` of an existing role or user adds to it. Global objects are nobody's own by being new. |\n| \"Revoking can only make it safer\" | `ALTER ROLE … REVOKE` on a role others hold cuts them off too, and a data source or a Scheduler job that logs in as the person stops working tonight. The yes is to the revokes, shown with who keeps what. |\n| \"I can't log in as them, so I'll create a test user\" | A user with a password puts it in a file and in the transcript; a user with a role is a person with access. Impersonate the real people, or report the check as not done. |\n| \"I'll give my profile the `impersonator` role to check\" | That changes your account's privileges on a shared server — the administrator's decision. Ask for it; meanwhile the result is unverified. |\n\n**Red flags — stop:** the database in `VIEW_DATABASES`, or a view you tag, existed before\nthis session and you are about to apply rather than show; `env.production` is `true`; you\nare about to run `ALTER USER`, `ALTER ROLE`, `CREATE OR REPLACE` of a role, a user, a tag a\npolicy names, or a policy that existed; you are writing `DROP TAG … CASCADE` or `DROP ROLE`;\nyou are about to call a restriction verified without an impersonated query; you are typing a\npassword.\n\n## Templates\n\nThe four blocks below run in this order: the role, then the tag and the policy, then the\nview's own file, then the check. Names follow the example database of `\u002Fdenodo:views`.\n\n### A role that reads the marts, given to a user\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCREATE OR REPLACE ROLE sales_analyst 'Sales analysts: read the household marts'\n    GRANT CONNECT ON sales_analytics\n    GRANT EXECUTE ON sales_analytics.household_income\n    GRANT EXECUTE ON sales_analytics.household_income_by_band;\n\nALTER USER mlee GRANT ROLE sales_analyst;\n```\n\n- **There is no `SELECT` privilege**: reading is `EXECUTE`, and nothing works without\n  `CONNECT` on the database. `EXECUTE` on a view is enough to query it — the views below it\n  need no grant, and are better left without one (see \"Where the tag goes\").\n- **Grant the views, not the database.** `GRANT EXECUTE ON sales_analytics` hands over every\n  base view, every view added later, and makes the role read past a tag placed on the top\n  view only.\n- **`CREATE OR REPLACE ROLE` adds, never removes**: re-declared with fewer `GRANT`s, a role\n  kept every privilege and every member — *verified: 9.5.1 (live, 2026-10-01)*. Deleting a\n  line from this file revokes nothing; revoking is `ALTER ROLE sales_analyst REVOKE EXECUTE\n  ON sales_analytics.household_income_by_band`.\n- `ALTER USER … GRANT ROLE` adds the role; the user's other roles stay. A user's privileges\n  are the union of their own and every role's — no role narrows another.\n- Giving a role to a person is always the second row of the table above.\n\n### Tag the columns, then the policy\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCREATE OR REPLACE TAG personal_data\n    DESCRIPTION = 'Personal data of a household. Masked for sales analysts by policy analyst_masking.';\n\nCREATE OR REPLACE GLOBAL_SECURITY_POLICY analyst_masking\n    DESCRIPTION = 'Sales analysts get the columns tagged personal_data masked: text as asterisks, numbers and dates empty.'\n    ENABLED = TRUE\n    AUDIENCE ( ANY ROLES ( sales_analyst ) )\n    ELEMENTS ( VIEW_DATABASES ( sales_analytics ) COLUMNS TAGGED ANY ( personal_data ) )\n    RESTRICTION ( FILTER = ''\n                  MASKING ANY ( personal_data )\n                  WITH ( HIDE ) ( texts WITH REDACT_ASTERISK, numbers WITH HIDE, datetimes WITH HIDE ) );\n```\n\n- The tag exists before anything names it — the policy (`The following tags do not exist`)\n  and the view's file below. The columns get it in their views' own files.\n- **`VIEW_DATABASES` always**: without it the policy is checked on the views of every\n  database of the server.\n- **Name every type the tagged columns have.** A column whose type is missing from the list\n  came back `NULL`. `HIDE` is `NULL` for any type; `ROUND` on an integer is no mask at all.\n  The other expressions and the row filter and deny forms are in `references\u002Fpolicies.md`:\n  a filter is `RESTRICTION ( FILTER = '\u003Ctag> = ''EMEA''' REJECT )` — the condition names the\n  tag, not the column — and a deny is `ELEMENTS ( … VIEWS TAGGED ANY ( \u003Ctag> ) )` with\n  `RESTRICTION ( DENY )`.\n- `ENABLED` and `FILTER = ''` are mandatory: without them the statement is a syntax error.\n- The audience decides which grants are restricted — next section, before you pick one.\n\n### The tag in the view's own file\n\nThe view of `\u002Fdenodo:views`, re-declared with its field properties. This is the file that is\nre-applied on every change of the view, so this is where the tag has to live:\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nCREATE OR REPLACE VIEW iv_household_income\n    FOLDER = '\u002F02 - integration'\n    DESCRIPTION = 'Households enriched with the bounds of their income band. One row per household.'\n    PRIMARY KEY ( 'household_sk' )\n    ( buy_potential TAGS ( personal_data ), dependents TAGS ( personal_data ) )\n    AS SELECT hd.hd_demo_sk         AS household_sk,\n              hd.hd_income_band_sk  AS income_band_sk,\n              hd.hd_buy_potential   AS buy_potential,\n              hd.hd_dep_count       AS dependents,\n              hd.hd_vehicle_count   AS vehicles,\n              ib.ib_lower_bound     AS income_lower_bound,\n              ib.ib_upper_bound     AS income_upper_bound\n       FROM bv_household_demographics hd\n            INNER JOIN bv_income_band ib\n            ON hd.hd_income_band_sk = ib.ib_income_band_sk\n    CONTEXT ('formatted' = 'yes');\n```\n\n- **Re-applying the view's file without the `TAGS` removes the tag, and the mask with it** —\n  for a derived view and for a base view alike, without an error — *verified: 9.5.1 (live,\n  2026-10-01)*. A base view carries it in its `CREATE TABLE` column list:\n  `hd_buy_potential:text TAGS (personal_data),`.\n- On a view whose file is not yours, the tag goes on with\n  `ALTER TAG personal_data ADD_TO ( VIEWS () COLUMNS ( sales_analytics.\u003Cview>.\u003Ccolumn> ) ) REMOVE_FROM ( VIEWS () COLUMNS () )`\n  (`\u002Fdenodo:catalog`; *verified: 9.5.1 (live, 2026-10-05)*) — after the yes — and the message says that the owner's file must get\n  the same `TAGS`, or their next apply unmasks it. `ALTER TAG` keeps the tag's description;\n  `CREATE OR REPLACE TAG … ADD_TO` does the same assignment but rewrites the description and\n  is not flagged by the tool at all — which changes nothing about the yes.\n- A mask applies where the tag is, and everything computed above sees the masked value\n  (`references\u002Fpolicies.md`, Masking).\n\n### Check it as the people it is for\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nSELECT buy_potential, dependents FROM household_income LIMIT 5\nCONTEXT ('impersonate_user' = 'mlee');\n\nSELECT COUNT(*) AS unmasked FROM household_income WHERE buy_potential \u003C> '********'\nCONTEXT ('impersonate_user' = 'mlee');\n\nSELECT avg_dependents FROM household_income_by_band LIMIT 3\nCONTEXT ('impersonate_roles' = 'sales_analyst,allusers');\n```\n\n- The impersonated queries are the first three rows of *Verify* below; write down what each\n  returned — that table is the result of the work.\n- A role nobody holds yet is checked with `impersonate_roles` — named roles only, so `allusers`,\n  which a local user holds too, goes with it. When no reader outside the audience exists, say so —\n  the administrator is not one: it sees everything anyway.\n- A file stops at its first failure: an expected refusal (`does not have EXECUTE\n  privileges`) goes last, or into a call of its own.\n- A filter compares the masked value, so `unmasked` is `0` when every row is masked. For a\n  row filter, compare `COUNT(*)` as the person with the admin's `COUNT(*) WHERE \u003Ccondition>`.\n- **Impersonation checks reads only.** An `INSERT`, `UPDATE` or `DELETE` with the same\n  `CONTEXT` ran with the profile's own privileges: a user with nothing but `EXECUTE` on a view\n  updated it and a base view they had no grant on — *verified: 9.5.1 (live, 2026-10-02)*.\n  Whether someone may write is proven only by a session as them — their own login, or (documentation\n  only) `CONNECT USER \u003Cu>` without a password from a profile with the `impersonator` role. Both run\n  real writes: hand that check to the human (`\u002Fdenodo:dml`).\n\n### Each person sees only their own rows\n\nThe rows carry the login of the person they belong to; a tag on that column, and a filter that\ncompares the tag with the login of whoever is asking:\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nCREATE OR REPLACE TAG row_owner\n    DESCRIPTION = 'The login of the person a row belongs to. Sales reps see only their own rows: policy own_opportunities_only.';\n\nCREATE OR REPLACE VIEW opportunities\n    FOLDER = '\u002F03 - business entities'\n    DESCRIPTION = 'Open sales opportunities with the login of the rep who owns each. One row per opportunity. Reps see only their own: policy own_opportunities_only.'\n    PRIMARY KEY ( 'opportunity_id' )\n    ( sales_rep TAGS ( row_owner ) )\n    AS SELECT opportunity_id, account_name, amount, sales_rep\n       FROM iv_opportunities\n    CONTEXT ('formatted' = 'yes');\n\nCREATE OR REPLACE GLOBAL_SECURITY_POLICY own_opportunities_only\n    DESCRIPTION = 'Sales reps see only the rows whose column tagged row_owner holds their own login, case ignored, in sales_analytics and every view built on those columns.'\n    ENABLED = TRUE\n    AUDIENCE ( ANY ROLES ( sales_reps ) )\n    ELEMENTS ( VIEW_DATABASES ( sales_analytics ) COLUMNS TAGGED ANY ( row_owner ) )\n    RESTRICTION ( FILTER = 'UPPER(row_owner) = UPPER(GETSESSION(''user''))' REJECT );\n```\n\n- **`GETSESSION('user')` in the condition is the login of whoever runs the query** — under\n  `impersonate_user`, the person impersonated. The condition names the tag, not the column.\n- **`COLUMNS TAGGED`, not `VIEWS TAGGED`.** With the tag on a column, `VIEWS TAGGED ANY ( row_owner\n  )` is created, `GET_ELEMENTS()` says `OK`, and nobody's rows are filtered.\n- **Before the policy, compare the column with the people**: `SELECT sales_rep, COUNT(*)\n  FROM opportunities GROUP BY sales_rep` against `GET_USERS_WITH_ROLE()` of the\n  audience. A login is case-sensitive (impersonating `MLEE` is `The user 'MLEE' does not exist`),\n  so a value stored in another case matches nobody under `=`; a row with no owner is seen by\n  nobody in the audience, nor is a departed person's. Show the human the values that match no\n  login. `UPPER` on both sides is a choice — it would also let a future user `MLEE` see `mlee`'s\n  rows — and the exact `=` is the other.\n- The audience sees its own rows in this view and in every view built on it — an aggregate above\n  counts only theirs. Everyone outside the audience, a director through another role, and every\n  administrator sees all rows (\"Choosing the audience\").\n- The tag lives in the view's file, as above (\"The tag in the view's own file\"); on someone else's\n  view it goes on with `ALTER TAG` after the yes.\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\n-- 1. What each person should see, as the administrator: rows per owner.\nSELECT UPPER(sales_rep) AS owner, COUNT(*) AS rows_owned\nFROM opportunities\nGROUP BY UPPER(sales_rep);\n\n-- 2. As each person of the audience: their count, and 0 rows of anyone else.\nSELECT COUNT(*) AS rows_seen,\n       SUM(CASE WHEN sales_rep IS NULL OR UPPER(sales_rep) \u003C> UPPER('mlee') THEN 1 ELSE 0 END) AS not_theirs\nFROM opportunities\nCONTEXT ('impersonate_user' = 'mlee');\n```\n\n**Check a per-person filter only with `impersonate_user`**, one query per person: under\n`impersonate_roles` the session's user is `__roles_impersonator_user__`, so the filter returns no\nrows — it looks like the restriction works and proves nothing (*verified: 9.5.1 (live,\n2026-10-06)*). `rows_seen` equals that person's `rows_owned`, and `not_theirs` is `0`.\n\n### Take someone's access away\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\n-- 1. Every path into the database: the rows with userrolename empty are granted to the person,\n--    the others come with a role.\nSELECT username, userrolename, elementname, dbconnect, dbexecute, elementexecute, dbadmin\nFROM CATALOG_PERMISSIONS()\nWHERE dbname = 'sales_analytics' AND username = 'kchen';\n\n-- 2. Who else holds each role the first query names: revoking from the role cuts them off too.\nSELECT name FROM GET_USERS_WITH_ROLE() WHERE role = 'sales_analyst' AND include_indirect_roles = true;\n\n-- 3. Who holds it directly: someone only in 2 has it through another role, and that is the role to revoke.\nSELECT name FROM GET_USERS_WITH_ROLE() WHERE role = 'sales_analyst' AND include_indirect_roles = false;\n```\n\nThe file, one statement per path of query 1 — it waits for the yes (Who applies what):\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nALTER USER kchen REVOKE ROLE sales_analyst;\nALTER USER kchen REVOKE EXECUTE ON sales_analytics.household_income;\nALTER USER kchen REVOKE CONNECT ON sales_analytics;\n```\n\n- **From the person, never from a role others hold**: `ALTER ROLE … REVOKE` takes it from every\n  member of query 2.\n- **Every row of query 1 is a way in, and each one alone reads data**: a direct `EXECUTE` on a view\n  kept working after the person's own `CONNECT` was revoked, because a role still granted\n  `CONNECT`; a role with `EXECUTE` on the whole database reads every view.\n- **`REVOKE` of a role or a privilege the person does not have answers `ok`** — a misspelled role\n  too. Query 1 again, after, is the answer: no row.\n- **The check as the person runs from another database** — the profile's own — naming the view\n  with its database: `SELECT COUNT(*) FROM sales_analytics.household_income CONTEXT\n  ('impersonate_user' = 'kchen')` answers `The user does not have CONNECT privileges on the\n  database 'sales_analytics'` — that database, not the one you run from. Run in\n  `sales_analytics` itself, an impersonated query does not check `CONNECT`: a person with nothing\n  left but an `EXECUTE` on one view still read it there — *verified: 9.5.1 (live, 2026-10-06)*.\n- **What a revoke does not reach**: `dbadmin = true` in query 1 — a local administrator — and a\n  global administrator (`LIST USERS`, column `admin`) read everything; a role that comes from an\n  LDAP or identity-provider group comes back at the next login — the administrator's, in Design\n  Studio or the identity provider; copies the person already exported.\n- **What stops working with it**: a VDP data source or a Scheduler job that logs in as the person\n  (`api --server scheduler get \u002Fpublic\u002Fapi\u002FdataSources` → `login`). Name them in the message.\n- The message is the shape above — readers today, the file, who reads what after, what is not\n  covered.\n\n## Choosing the audience\n\n**An audience restricts only the grant path it names; every other path that grants\n`EXECUTE` on the view returns the data unrestricted** — *verified: 9.5.1 (live, 2026-10-01)*:\n\n| Audience | Restricted | Reads in clear anyway |\n|---|---|---|\n| `ANY ROLES ( r )` | what role `r` grants | a direct grant to the user (on the database **or on that one view**), any other role that grants `EXECUTE`, a local administrator |\n| `NOT_IN ROLES ( r )` | what every role except `r` grants | a direct grant to the user, a local administrator |\n| `ANY USERS ( u )`, `NOT_IN USERS ( u )` | what is granted to those users directly | everything that comes through their roles, a local administrator |\n| `ALL` | every path, every non-administrator | local and global administrators only |\n\nSo: before a role audience, read query 1 — every path in the third column is a person who\nkeeps reading in clear. One audience takes one form — roles or users, never both\n(`Syntax error … near 'USERS'`) — so the two ways to close a direct grant are revoking it\n(`ALTER USER kchen REVOKE EXECUTE ON sales_analytics.household_income`) or a second policy,\nthe same but with `AUDIENCE ( ANY USERS ( kchen ) )` — either way the user was masked from\nthen on, *verified: 9.5.1 (live, 2026-10-01)*. Both are the human's choice, with a yes.\n\"Everyone except the auditors\" with `ALL` is impossible — `ALL` restricts them too;\n`NOT_IN ROLES ( auditor )` works only while nobody reads by a direct grant.\n\n## Where the tag goes\n\n- A mask or a filter on a tagged column applies to that view and **to every view built on\n  it**; an aggregate over the column changes with it.\n- A tag on the top view only leaves the views below it unrestricted. Tag the lowest view that\n  carries the column — the base view when the audience can read the whole database — or grant\n  the audience only the views above the tag.\n- \"It must also cover views built later\": a tag on the base view's column covers every view\n  built over it; a new source with its own copy of the column needs its own tag.\n- **Many columns** — \"mask every column that holds an email\": the list, its lineage, where each\n  assignment goes and the check row for row are `\u002Fdenodo:catalog`, *One tag on many columns*;\n  which rows wait, *Who applies what*. The impersonated check runs per view of the list, not on\n  one example.\n\n## What you need\n\n| Slot | Where it comes from |\n|---|---|\n| Who is restricted, who is not | the human, as people and roles; then query 1 — the grant paths decide the audience |\n| Which columns, which views | the human names the data, often by another name than the column (`buy potential` is `hd_buy_potential` two views down): `COLUMN_DEPENDENCIES()` of the view they read gives the base column. Columns of the same kind the human did not name go into the message, not into the tag |\n| How each is shown | the human: masked text, empty, a filter, no access. Not said → ask; the treatment of numbers and dates is part of the question |\n| Which databases | the database of every view that carries the tag — the base view's own when the tag sits there — not only the one people query; always in `VIEW_DATABASES` |\n| Tag name | the concept (`personal_data`, `sales_territory`); check that no policy already names it (query 3) |\n| Can you do it, and check it | `env check` → `features.enterprise_plus` (`false`: tags and policies need Enterprise Plus — say so and stop); `vdp.admin` (`false`: a role needs `create_role`, assigning it `assign_all_roles`, a privilege `ADMIN` on the database or `assignprivileges`, a tag `manage_tags`, a policy `ADMIN` on each database it names or `manage_policies` — else the files go to the human); `vdp.impersonation` |\n\n## Verify\n\n| Check | Query | Expect |\n|---|---|---|\n| Each person in the audience | the column `SELECT … CONTEXT ('impersonate_user' = '\u003Cuser>')` | masked \u002F only their rows \u002F `does not have privileges to execute view` |\n| A person outside it | the same, as them | the real values |\n| Views below the tagged one | the same query on each view the audience can read | restricted too, or not readable |\n| The policy is evaluated | query 2 | `element_status = 'OK'` |\n| The tags landed | query 4 | one row per tagged column — a misspelled column is accepted silently (`\u002Fdenodo:catalog`) |\n| No path left over | query 1 | nothing outside the audience reads a tagged view, or the human was told |\n\n## Silent failures\n\nEach runs without an error — *verified: 9.5.1 (live, 2026-10-01)*; rows 13–16 —\n*verified: 9.5.1 (live, 2026-10-06)*.\n\n| You did | What happens | Instead |\n|---|---|---|\n| 1. checked as yourself | an administrator sees everything | impersonate |\n| 2. a role audience, and someone also has a direct grant or another role with `EXECUTE` | that person reads in clear | read query 1, name them |\n| 3. a user audience for people who read through roles | they read in clear | a role audience, or `ALL` |\n| 4. tagged only the top view | the views below read in clear for whoever may read them | tag the lowest view; grant views, not the database |\n| 5. re-applied a view's file without its `TAGS` | the tag is gone, and the mask | the tags live in the view's file |\n| 6. `DROP ROLE` of a role named in a policy's audience | the policy stops being evaluated for **every** role in it; `DESC VQL` still says `ENABLED = TRUE` and ends with `# Invalid object policy`; query 2 says `INVALID` | read query 2 and the audiences before dropping a role. Afterwards two repairs, both a yes: re-create the role empty — the policy is validated and comes back unchanged — or `CREATE OR REPLACE` the policy without it. Which one depends on why the role went: ask |\n| 7. `DROP TAG … CASCADE` | the tag and every policy that names it are deleted (without `CASCADE`: `Some elements depend on`) | query 3 first |\n| 8. removed a `GRANT` line from a role's file and re-applied it | nothing revoked | `ALTER ROLE … REVOKE` — a yes |\n| 9. left a type out of the masking list | that column comes back `NULL` | name texts, numbers and dates |\n| 10. masked a column people group, filter or join on | groups collapse into `********`, filters find nothing | say it in the message |\n| 11. `ALTER DATABASE … GRANT … TO ROLE` with a misspelled role | `ok`, nothing granted | read query 1 back |\n| 12. trusted `GET_CATALOG_EFFECTIVE_PERMISSIONS` | `rowpermissions = true` for a user who reads in clear | impersonate |\n| 13. a row filter with `VIEWS TAGGED` while the tag is on a column | created, `OK`, filters nothing | `COLUMNS TAGGED` |\n| 14. a per-person filter checked with `impersonate_roles` | no rows — the session's user is `__roles_impersonator_user__` | `impersonate_user`, per person |\n| 15. checked a revoke as the person inside the database itself | `CONNECT` is not checked there: a person with one `EXECUTE` left still read | the impersonated query from another database, `\u003Cdb>.\u003Cview>` |\n| 16. `ALTER USER … REVOKE` of a role or privilege the person does not have | `ok` | `CATALOG_PERMISSIONS()` for the person, after |\n\n## Common mistakes\n\n| You wrote | Server says | Fix |\n|---|---|---|\n| a policy without `ENABLED` | `Syntax error: Exception parsing query near 'AUDIENCE'` | `ENABLED = TRUE` |\n| `MASKING` without `FILTER = ''` | `Syntax error: Exception parsing query near 'MASKING'` | `FILTER = '' MASKING …` |\n| `MASKING ANY ( t )` with no `WITH` | `Syntax error: Cannot invoke \"…getDefaultMasking()\" because \"mt\" is null` | `WITH ( HIDE ) ( texts WITH …, numbers WITH …, datetimes WITH … )` |\n| an audience role that does not exist | `There was an error creating the global security policy. Role 'x' does not exist` | the role first; check the spelling with `--max-rows` |\n| a tag that does not exist | `The following tags do not exist: 'x'` | `CREATE OR REPLACE TAG` first |\n| `CREATE USER u EXTERNAL GRANT ROLE r` | `Syntax error … near 'GRANT'` | accounts are the administrator's; a role is given with `ALTER USER u GRANT ROLE r` |\n| an impersonated query | `This user cannot impersonate. Only users with role 'impersonator' can impersonate.` | the check is unverified — hand the human the queries and the expected answers; ask the administrator |\n| an impersonated query | `The user does not have CONNECT privileges on the database '\u003Cdb>'` | a real answer: that user cannot read the database |\n| `GET_USERS_WITH_ROLE()` with only `role` | `No search methods ready to be run. … include_indirect_roles` | `WHERE role = 'r' AND include_indirect_roles = true` |\n\n## Reference\n\n- `references\u002Fpolicies.md` — the `CREATE GLOBAL_SECURITY_POLICY` grammar, every masking\n  expression as measured, the row filter and deny, enabling and disabling, the audience\n  table per grant path, what removes or disables a policy silently, and its errors.\n- `references\u002Fprivileges.md` — roles, users and grants (`CREATE`\u002F`ALTER ROLE`, `ALTER USER`,\n  `ALTER DATABASE … TO`), the privilege list, column privileges, owners (`CHOWN`), the\n  permission procedures and what each one proves, and impersonation.\n","# Who may read what\n\nEverything here is server-wide, and it changes what real people get back from views —\nsilently. The statement succeeds; then an analyst's report shows `********` where a name\nwas, an amount comes back `NULL`, one analyst still reads every row, or a policy stops\nbeing evaluated without anyone being told. Two facts shape every step below:\n\n1. **Administrators are never restricted** — global administrators, and local administrators of\n   the view's database (`ADMIN` on it). The profile you connect with may be one: `env check` →\n   `vdp.admin`. **Your own `SELECT` proves nothing about a restriction.**\n2. **The proof is the same query run as the person** — `CONTEXT ('impersonate_user' = '\u003Cuser>')`,\n   or `('impersonate_roles' = '\u003Crole>')` — with their privileges and every policy, and no password.\n   It needs `vdp.impersonation: true` (the profile's user has the `impersonator` role).\n\nA tag as a plain label is `\u002Fdenodo:catalog` (its full syntax lives there); descriptions and the\nMCP visibility tag are `\u002Fdenodo:semantics`. User accounts and passwords, LDAP and\nidentity-provider groups, per-role row and column restrictions, custom policies,\nsession-attribute audiences and server properties are set in Design Studio by the\nadministrator — say so and stop. Applying files is `\u002Fdenodo:execute`; the working loop and the\ncore safety rule are `\u002Fdenodo:vql`.\n\n## Before you change anything\n\nReads only, and every decision below comes from them — complete only for a global administrator\n(`vdp.admin: true`). For anyone else, a local administrator too, query 1 silently returns only the\nprofile's own grants and its roles': say the list of readers is partial.\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\n-- 1. Who reads the database, and through what: one row per grant. A user's row with\n--    userrolename empty = granted to that user directly; a row with username empty = a\n--    role's own grant (a role nobody holds shows only this way); elementname empty = the\n--    whole database\nSELECT username, userrolename, elementname, dbconnect, dbexecute, elementexecute, dbadmin\nFROM CATALOG_PERMISSIONS() WHERE dbname = 'sales_analytics';\n\n-- 2. Policies already on the server, and whether each is still evaluated (INVALID = not)\nSELECT name, element_status, description FROM GET_ELEMENTS() WHERE type = 'globalSecurityPolicy';\n\n-- 3. Which policy reads which tag\nSELECT global_security_policy_name, tag_name FROM GET_GLOBAL_SECURITY_POLICIES_TAGS();\n\n-- 4. Tags already on the views of the database\nSELECT view_name, column_name, tag_name FROM GET_VIEW_TAGS() WHERE input_database_name = 'sales_analytics';\n```\n\n- Then `DESC VQL GLOBAL_SECURITY_POLICY \u003Cname>` for each policy that names a tag, a role or a\n  database you are about to touch — its audience is not in any of the queries above. Like query 3,\n  it needs an administrator of the database. It starts with `DROP … CASCADE`: read it, never apply it.\n- **A tag that a policy names is a security switch**, not a label: adding it to a column\n  restricts that column for the policy's audience, removing it lifts the restriction.\n- One user's grants in the database: query 1 with `AND username = '\u003Cuser>'`. Not `DESC VQL USER`:\n  without `('includeUserPrivileges' = 'yes')` it shows no roles at all, and with it, for a\n  local user, it prints the password hash into the transcript.\n- Someone known only to LDAP, Kerberos or an identity provider is no VDP user: no row in query 1, and\n  `impersonate_user` says `does not exist`. Their access is the roles of their groups: ask the human\n  which, check them with `impersonate_roles`, and hand the check of a per-person filter to the human.\n- Whether a role exists: `DESC ROLE \u003Cname>` (`Error loading role` when it does not).\n  `GET_USERS_WITH_ROLE()` answers a missing role with 0 rows and no error, and `LIST ROLES`\n  is cut at the tool's 100 rows (`truncated: true`) unless you pass `--max-rows 5000`.\n\n**A user sees what they should not** — in this order, each step a read:\n\n1. Query 2: is the policy `INVALID`? `DESC VQL` of it ends with `# Invalid object policy`;\n   then `DESC ROLE` each role of its audience — one that is gone switched the policy off for\n   all of them (Silent failures, 6).\n2. Query 1 for that user: a row with `userrolename` empty, a role outside the audience, or\n   `dbadmin = true` is a path the policy does not restrict (Choosing the audience).\n3. Query 4: is the tag on the lowest view the user can read, and on the column the view\n   exposes? `SELECT column_name, dependency_name, depth FROM COLUMN_DEPENDENCIES() WHERE\n   input_view_database_name = '\u003Cdb>' AND input_view_name = '\u003Cview>'` names the views and the\n   base column behind each output column (`\u002Fdenodo:views`, lineage) — *verified: 9.5.1 (live, 2026-10-05)*.\n4. The impersonated query as that user, now and after each fix you propose.\n\n## Who applies what\n\n| What the statements touch | You apply yourself | Only after the human's yes |\n|---|---|---|\n| only objects you created in this session: a role no person holds yet, whose grants name only objects you created in this session, a policy limited by `VIEW_DATABASES` to a database you created in this session and whose audience is such a role, tags on views you created | create, change, check | — |\n| anything that existed before this session — a role, a user, a tag a policy names, a policy, someone else's view; any grant or revoke of a role or a privilege to a person; a new policy whose audience or views include existing ones; re-creating a role someone dropped | the reads, the impersonated checks, the files | every statement |\n\n`vql plan` on the file names, per statement, every existing object it touches that this\nsession did not create (`touches`) — the first row needs that list empty (`\u002Fdenodo:vql`).\nThe tool marks `CREATE` of a role, a user or a policy `destructive: security`.\n\nThe yes is to the statements, after you have shown them in this shape:\n\n```\nDatabase sales_analytics. Readers today (CATALOG_PERMISSIONS): role sales_analyst — mlee,\nrpatel; kchen also has EXECUTE on household_income granted to her directly.\n1. security\u002Fpersonal_data.vql: tag personal_data, policy analyst_masking (role sales_analyst)\n2. model\u002Fiv_household_income.vql: the view's file gains TAGS on buy_potential, dependents\nAfter: mlee, rpatel see buy_potential as ********, dependents as NULL — in every view on\niv_household_income; avg_dependents in household_income_by_band becomes NULL for them.\nNot covered: kchen still reads both in clear (direct grant). Revoking it is a separate yes.\nAlso exposed, not asked about: the role reads every column of the base views below.\nApply 1 and 2?\n```\n\nWhen you cannot ask — the human is away, the deadline is close — the answer is the files and\nthis message, with the impersonated checks the human can run after applying, not the\nstatements applied. Say what waiting costs, and which file goes first. A file waiting for a\nyes keeps the template's `verified:` mark — the template was run, this file was not — and\nthe message says it is not applied. It waits whole: a statement the plan calls yours (a new tag)\nwaits with the rest, or it is left behind with nothing to do. Run its impersonated checks now anyway: today's answers\nare the \"before\" the human compares with.\n\n| Rationalization | Reality |\n|---|---|\n| \"They named the exact columns and the treatment — that is the yes\" \u002F \"they said *set it up and make sure it works*\" | They named an outcome. They have not seen that a number becomes `NULL` in every report, nor which user the policy leaves out. The yes comes after the statements. |\n| \"Every object is new, carries my prefix and is limited to one database\" | A name you chose does not make what it touches yours. The policy restricts the people who read that database, the tag sits on someone else's view: the first row of the table needs the database and the views created by you in this session, not the policy. |\n| \"It only restricts — it can only make things safer\" | A mask changes what reports compute: a `GROUP BY` on the masked column collapses into one group, an aggregate over a hidden number becomes `NULL`, a filter on it finds nothing. Someone's dashboard breaks silently in the morning. |\n| \"It is reversed with one `REMOVE_FROM` \u002F `DROP`\" | The readers already got the result; and `DROP TAG … CASCADE` deletes every policy that names the tag. |\n| \"Only new objects — a pure `CREATE`\" | A new policy restricts existing people; a re-created role revives an invalid policy; `CREATE OR REPLACE` of an existing role or user adds to it. Global objects are nobody's own by being new. |\n| \"Revoking can only make it safer\" | `ALTER ROLE … REVOKE` on a role others hold cuts them off too, and a data source or a Scheduler job that logs in as the person stops working tonight. The yes is to the revokes, shown with who keeps what. |\n| \"I can't log in as them, so I'll create a test user\" | A user with a password puts it in a file and in the transcript; a user with a role is a person with access. Impersonate the real people, or report the check as not done. |\n| \"I'll give my profile the `impersonator` role to check\" | That changes your account's privileges on a shared server — the administrator's decision. Ask for it; meanwhile the result is unverified. |\n\n**Red flags — stop:** the database in `VIEW_DATABASES`, or a view you tag, existed before\nthis session and you are about to apply rather than show; `env.production` is `true`; you\nare about to run `ALTER USER`, `ALTER ROLE`, `CREATE OR REPLACE` of a role, a user, a tag a\npolicy names, or a policy that existed; you are writing `DROP TAG … CASCADE` or `DROP ROLE`;\nyou are about to call a restriction verified without an impersonated query; you are typing a\npassword.\n\n## Templates\n\nThe four blocks below run in this order: the role, then the tag and the policy, then the\nview's own file, then the check. Names follow the example database of `\u002Fdenodo:views`.\n\n### A role that reads the marts, given to a user\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCREATE OR REPLACE ROLE sales_analyst 'Sales analysts: read the household marts'\n    GRANT CONNECT ON sales_analytics\n    GRANT EXECUTE ON sales_analytics.household_income\n    GRANT EXECUTE ON sales_analytics.household_income_by_band;\n\nALTER USER mlee GRANT ROLE sales_analyst;\n```\n\n- **There is no `SELECT` privilege**: reading is `EXECUTE`, and nothing works without\n  `CONNECT` on the database. `EXECUTE` on a view is enough to query it — the views below it\n  need no grant, and are better left without one (see \"Where the tag goes\").\n- **Grant the views, not the database.** `GRANT EXECUTE ON sales_analytics` hands over every\n  base view, every view added later, and makes the role read past a tag placed on the top\n  view only.\n- **`CREATE OR REPLACE ROLE` adds, never removes**: re-declared with fewer `GRANT`s, a role\n  kept every privilege and every member — *verified: 9.5.1 (live, 2026-10-01)*. Deleting a\n  line from this file revokes nothing; revoking is `ALTER ROLE sales_analyst REVOKE EXECUTE\n  ON sales_analytics.household_income_by_band`.\n- `ALTER USER … GRANT ROLE` adds the role; the user's other roles stay. A user's privileges\n  are the union of their own and every role's — no role narrows another.\n- Giving a role to a person is always the second row of the table above.\n\n### Tag the columns, then the policy\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCREATE OR REPLACE TAG personal_data\n    DESCRIPTION = 'Personal data of a household. Masked for sales analysts by policy analyst_masking.';\n\nCREATE OR REPLACE GLOBAL_SECURITY_POLICY analyst_masking\n    DESCRIPTION = 'Sales analysts get the columns tagged personal_data masked: text as asterisks, numbers and dates empty.'\n    ENABLED = TRUE\n    AUDIENCE ( ANY ROLES ( sales_analyst ) )\n    ELEMENTS ( VIEW_DATABASES ( sales_analytics ) COLUMNS TAGGED ANY ( personal_data ) )\n    RESTRICTION ( FILTER = ''\n                  MASKING ANY ( personal_data )\n                  WITH ( HIDE ) ( texts WITH REDACT_ASTERISK, numbers WITH HIDE, datetimes WITH HIDE ) );\n```\n\n- The tag exists before anything names it — the policy (`The following tags do not exist`)\n  and the view's file below. The columns get it in their views' own files.\n- **`VIEW_DATABASES` always**: without it the policy is checked on the views of every\n  database of the server.\n- **Name every type the tagged columns have.** A column whose type is missing from the list\n  came back `NULL`. `HIDE` is `NULL` for any type; `ROUND` on an integer is no mask at all.\n  The other expressions and the row filter and deny forms are in `references\u002Fpolicies.md`:\n  a filter is `RESTRICTION ( FILTER = '\u003Ctag> = ''EMEA''' REJECT )` — the condition names the\n  tag, not the column — and a deny is `ELEMENTS ( … VIEWS TAGGED ANY ( \u003Ctag> ) )` with\n  `RESTRICTION ( DENY )`.\n- `ENABLED` and `FILTER = ''` are mandatory: without them the statement is a syntax error.\n- The audience decides which grants are restricted — next section, before you pick one.\n\n### The tag in the view's own file\n\nThe view of `\u002Fdenodo:views`, re-declared with its field properties. This is the file that is\nre-applied on every change of the view, so this is where the tag has to live:\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nCREATE OR REPLACE VIEW iv_household_income\n    FOLDER = '\u002F02 - integration'\n    DESCRIPTION = 'Households enriched with the bounds of their income band. One row per household.'\n    PRIMARY KEY ( 'household_sk' )\n    ( buy_potential TAGS ( personal_data ), dependents TAGS ( personal_data ) )\n    AS SELECT hd.hd_demo_sk         AS household_sk,\n              hd.hd_income_band_sk  AS income_band_sk,\n              hd.hd_buy_potential   AS buy_potential,\n              hd.hd_dep_count       AS dependents,\n              hd.hd_vehicle_count   AS vehicles,\n              ib.ib_lower_bound     AS income_lower_bound,\n              ib.ib_upper_bound     AS income_upper_bound\n       FROM bv_household_demographics hd\n            INNER JOIN bv_income_band ib\n            ON hd.hd_income_band_sk = ib.ib_income_band_sk\n    CONTEXT ('formatted' = 'yes');\n```\n\n- **Re-applying the view's file without the `TAGS` removes the tag, and the mask with it** —\n  for a derived view and for a base view alike, without an error — *verified: 9.5.1 (live,\n  2026-10-01)*. A base view carries it in its `CREATE TABLE` column list:\n  `hd_buy_potential:text TAGS (personal_data),`.\n- On a view whose file is not yours, the tag goes on with\n  `ALTER TAG personal_data ADD_TO ( VIEWS () COLUMNS ( sales_analytics.\u003Cview>.\u003Ccolumn> ) ) REMOVE_FROM ( VIEWS () COLUMNS () )`\n  (`\u002Fdenodo:catalog`; *verified: 9.5.1 (live, 2026-10-05)*) — after the yes — and the message says that the owner's file must get\n  the same `TAGS`, or their next apply unmasks it. `ALTER TAG` keeps the tag's description;\n  `CREATE OR REPLACE TAG … ADD_TO` does the same assignment but rewrites the description and\n  is not flagged by the tool at all — which changes nothing about the yes.\n- A mask applies where the tag is, and everything computed above sees the masked value\n  (`references\u002Fpolicies.md`, Masking).\n\n### Check it as the people it is for\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nSELECT buy_potential, dependents FROM household_income LIMIT 5\nCONTEXT ('impersonate_user' = 'mlee');\n\nSELECT COUNT(*) AS unmasked FROM household_income WHERE buy_potential \u003C> '********'\nCONTEXT ('impersonate_user' = 'mlee');\n\nSELECT avg_dependents FROM household_income_by_band LIMIT 3\nCONTEXT ('impersonate_roles' = 'sales_analyst,allusers');\n```\n\n- The impersonated queries are the first three rows of *Verify* below; write down what each\n  returned — that table is the result of the work.\n- A role nobody holds yet is checked with `impersonate_roles` — named roles only, so `allusers`,\n  which a local user holds too, goes with it. When no reader outside the audience exists, say so —\n  the administrator is not one: it sees everything anyway.\n- A file stops at its first failure: an expected refusal (`does not have EXECUTE\n  privileges`) goes last, or into a call of its own.\n- A filter compares the masked value, so `unmasked` is `0` when every row is masked. For a\n  row filter, compare `COUNT(*)` as the person with the admin's `COUNT(*) WHERE \u003Ccondition>`.\n- **Impersonation checks reads only.** An `INSERT`, `UPDATE` or `DELETE` with the same\n  `CONTEXT` ran with the profile's own privileges: a user with nothing but `EXECUTE` on a view\n  updated it and a base view they had no grant on — *verified: 9.5.1 (live, 2026-10-02)*.\n  Whether someone may write is proven only by a session as them — their own login, or (documentation\n  only) `CONNECT USER \u003Cu>` without a password from a profile with the `impersonator` role. Both run\n  real writes: hand that check to the human (`\u002Fdenodo:dml`).\n\n### Each person sees only their own rows\n\nThe rows carry the login of the person they belong to; a tag on that column, and a filter that\ncompares the tag with the login of whoever is asking:\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\nCREATE OR REPLACE TAG row_owner\n    DESCRIPTION = 'The login of the person a row belongs to. Sales reps see only their own rows: policy own_opportunities_only.';\n\nCREATE OR REPLACE VIEW opportunities\n    FOLDER = '\u002F03 - business entities'\n    DESCRIPTION = 'Open sales opportunities with the login of the rep who owns each. One row per opportunity. Reps see only their own: policy own_opportunities_only.'\n    PRIMARY KEY ( 'opportunity_id' )\n    ( sales_rep TAGS ( row_owner ) )\n    AS SELECT opportunity_id, account_name, amount, sales_rep\n       FROM iv_opportunities\n    CONTEXT ('formatted' = 'yes');\n\nCREATE OR REPLACE GLOBAL_SECURITY_POLICY own_opportunities_only\n    DESCRIPTION = 'Sales reps see only the rows whose column tagged row_owner holds their own login, case ignored, in sales_analytics and every view built on those columns.'\n    ENABLED = TRUE\n    AUDIENCE ( ANY ROLES ( sales_reps ) )\n    ELEMENTS ( VIEW_DATABASES ( sales_analytics ) COLUMNS TAGGED ANY ( row_owner ) )\n    RESTRICTION ( FILTER = 'UPPER(row_owner) = UPPER(GETSESSION(''user''))' REJECT );\n```\n\n- **`GETSESSION('user')` in the condition is the login of whoever runs the query** — under\n  `impersonate_user`, the person impersonated. The condition names the tag, not the column.\n- **`COLUMNS TAGGED`, not `VIEWS TAGGED`.** With the tag on a column, `VIEWS TAGGED ANY ( row_owner\n  )` is created, `GET_ELEMENTS()` says `OK`, and nobody's rows are filtered.\n- **Before the policy, compare the column with the people**: `SELECT sales_rep, COUNT(*)\n  FROM opportunities GROUP BY sales_rep` against `GET_USERS_WITH_ROLE()` of the\n  audience. A login is case-sensitive (impersonating `MLEE` is `The user 'MLEE' does not exist`),\n  so a value stored in another case matches nobody under `=`; a row with no owner is seen by\n  nobody in the audience, nor is a departed person's. Show the human the values that match no\n  login. `UPPER` on both sides is a choice — it would also let a future user `MLEE` see `mlee`'s\n  rows — and the exact `=` is the other.\n- The audience sees its own rows in this view and in every view built on it — an aggregate above\n  counts only theirs. Everyone outside the audience, a director through another role, and every\n  administrator sees all rows (\"Choosing the audience\").\n- The tag lives in the view's file, as above (\"The tag in the view's own file\"); on someone else's\n  view it goes on with `ALTER TAG` after the yes.\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nCONNECT DATABASE sales_analytics;\n\n-- 1. What each person should see, as the administrator: rows per owner.\nSELECT UPPER(sales_rep) AS owner, COUNT(*) AS rows_owned\nFROM opportunities\nGROUP BY UPPER(sales_rep);\n\n-- 2. As each person of the audience: their count, and 0 rows of anyone else.\nSELECT COUNT(*) AS rows_seen,\n       SUM(CASE WHEN sales_rep IS NULL OR UPPER(sales_rep) \u003C> UPPER('mlee') THEN 1 ELSE 0 END) AS not_theirs\nFROM opportunities\nCONTEXT ('impersonate_user' = 'mlee');\n```\n\n**Check a per-person filter only with `impersonate_user`**, one query per person: under\n`impersonate_roles` the session's user is `__roles_impersonator_user__`, so the filter returns no\nrows — it looks like the restriction works and proves nothing (*verified: 9.5.1 (live,\n2026-10-06)*). `rows_seen` equals that person's `rows_owned`, and `not_theirs` is `0`.\n\n### Take someone's access away\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\n-- 1. Every path into the database: the rows with userrolename empty are granted to the person,\n--    the others come with a role.\nSELECT username, userrolename, elementname, dbconnect, dbexecute, elementexecute, dbadmin\nFROM CATALOG_PERMISSIONS()\nWHERE dbname = 'sales_analytics' AND username = 'kchen';\n\n-- 2. Who else holds each role the first query names: revoking from the role cuts them off too.\nSELECT name FROM GET_USERS_WITH_ROLE() WHERE role = 'sales_analyst' AND include_indirect_roles = true;\n\n-- 3. Who holds it directly: someone only in 2 has it through another role, and that is the role to revoke.\nSELECT name FROM GET_USERS_WITH_ROLE() WHERE role = 'sales_analyst' AND include_indirect_roles = false;\n```\n\nThe file, one statement per path of query 1 — it waits for the yes (Who applies what):\n\n```sql\n-- verified: 9.5.1 (live, 2026-10-07)\nALTER USER kchen REVOKE ROLE sales_analyst;\nALTER USER kchen REVOKE EXECUTE ON sales_analytics.household_income;\nALTER USER kchen REVOKE CONNECT ON sales_analytics;\n```\n\n- **From the person, never from a role others hold**: `ALTER ROLE … REVOKE` takes it from every\n  member of query 2.\n- **Every row of query 1 is a way in, and each one alone reads data**: a direct `EXECUTE` on a view\n  kept working after the person's own `CONNECT` was revoked, because a role still granted\n  `CONNECT`; a role with `EXECUTE` on the whole database reads every view.\n- **`REVOKE` of a role or a privilege the person does not have answers `ok`** — a misspelled role\n  too. Query 1 again, after, is the answer: no row.\n- **The check as the person runs from another database** — the profile's own — naming the view\n  with its database: `SELECT COUNT(*) FROM sales_analytics.household_income CONTEXT\n  ('impersonate_user' = 'kchen')` answers `The user does not have CONNECT privileges on the\n  database 'sales_analytics'` — that database, not the one you run from. Run in\n  `sales_analytics` itself, an impersonated query does not check `CONNECT`: a person with nothing\n  left but an `EXECUTE` on one view still read it there — *verified: 9.5.1 (live, 2026-10-06)*.\n- **What a revoke does not reach**: `dbadmin = true` in query 1 — a local administrator — and a\n  global administrator (`LIST USERS`, column `admin`) read everything; a role that comes from an\n  LDAP or identity-provider group comes back at the next login — the administrator's, in Design\n  Studio or the identity provider; copies the person already exported.\n- **What stops working with it**: a VDP data source or a Scheduler job that logs in as the person\n  (`api --server scheduler get \u002Fpublic\u002Fapi\u002FdataSources` → `login`). Name them in the message.\n- The message is the shape above — readers today, the file, who reads what after, what is not\n  covered.\n\n## Choosing the audience\n\n**An audience restricts only the grant path it names; every other path that grants\n`EXECUTE` on the view returns the data unrestricted** — *verified: 9.5.1 (live, 2026-10-01)*:\n\n| Audience | Restricted | Reads in clear anyway |\n|---|---|---|\n| `ANY ROLES ( r )` | what role `r` grants | a direct grant to the user (on the database **or on that one view**), any other role that grants `EXECUTE`, a local administrator |\n| `NOT_IN ROLES ( r )` | what every role except `r` grants | a direct grant to the user, a local administrator |\n| `ANY USERS ( u )`, `NOT_IN USERS ( u )` | what is granted to those users directly | everything that comes through their roles, a local administrator |\n| `ALL` | every path, every non-administrator | local and global administrators only |\n\nSo: before a role audience, read query 1 — every path in the third column is a person who\nkeeps reading in clear. One audience takes one form — roles or users, never both\n(`Syntax error … near 'USERS'`) — so the two ways to close a direct grant are revoking it\n(`ALTER USER kchen REVOKE EXECUTE ON sales_analytics.household_income`) or a second policy,\nthe same but with `AUDIENCE ( ANY USERS ( kchen ) )` — either way the user was masked from\nthen on, *verified: 9.5.1 (live, 2026-10-01)*. Both are the human's choice, with a yes.\n\"Everyone except the auditors\" with `ALL` is impossible — `ALL` restricts them too;\n`NOT_IN ROLES ( auditor )` works only while nobody reads by a direct grant.\n\n## Where the tag goes\n\n- A mask or a filter on a tagged column applies to that view and **to every view built on\n  it**; an aggregate over the column changes with it.\n- A tag on the top view only leaves the views below it unrestricted. Tag the lowest view that\n  carries the column — the base view when the audience can read the whole database — or grant\n  the audience only the views above the tag.\n- \"It must also cover views built later\": a tag on the base view's column covers every view\n  built over it; a new source with its own copy of the column needs its own tag.\n- **Many columns** — \"mask every column that holds an email\": the list, its lineage, where each\n  assignment goes and the check row for row are `\u002Fdenodo:catalog`, *One tag on many columns*;\n  which rows wait, *Who applies what*. The impersonated check runs per view of the list, not on\n  one example.\n\n## What you need\n\n| Slot | Where it comes from |\n|---|---|\n| Who is restricted, who is not | the human, as people and roles; then query 1 — the grant paths decide the audience |\n| Which columns, which views | the human names the data, often by another name than the column (`buy potential` is `hd_buy_potential` two views down): `COLUMN_DEPENDENCIES()` of the view they read gives the base column. Columns of the same kind the human did not name go into the message, not into the tag |\n| How each is shown | the human: masked text, empty, a filter, no access. Not said → ask; the treatment of numbers and dates is part of the question |\n| Which databases | the database of every view that carries the tag — the base view's own when the tag sits there — not only the one people query; always in `VIEW_DATABASES` |\n| Tag name | the concept (`personal_data`, `sales_territory`); check that no policy already names it (query 3) |\n| Can you do it, and check it | `env check` → `features.enterprise_plus` (`false`: tags and policies need Enterprise Plus — say so and stop); `vdp.admin` (`false`: a role needs `create_role`, assigning it `assign_all_roles`, a privilege `ADMIN` on the database or `assignprivileges`, a tag `manage_tags`, a policy `ADMIN` on each database it names or `manage_policies` — else the files go to the human); `vdp.impersonation` |\n\n## Verify\n\n| Check | Query | Expect |\n|---|---|---|\n| Each person in the audience | the column `SELECT … CONTEXT ('impersonate_user' = '\u003Cuser>')` | masked \u002F only their rows \u002F `does not have privileges to execute view` |\n| A person outside it | the same, as them | the real values |\n| Views below the tagged one | the same query on each view the audience can read | restricted too, or not readable |\n| The policy is evaluated | query 2 | `element_status = 'OK'` |\n| The tags landed | query 4 | one row per tagged column — a misspelled column is accepted silently (`\u002Fdenodo:catalog`) |\n| No path left over | query 1 | nothing outside the audience reads a tagged view, or the human was told |\n\n## Silent failures\n\nEach runs without an error — *verified: 9.5.1 (live, 2026-10-01)*; rows 13–16 —\n*verified: 9.5.1 (live, 2026-10-06)*.\n\n| You did | What happens | Instead |\n|---|---|---|\n| 1. checked as yourself | an administrator sees everything | impersonate |\n| 2. a role audience, and someone also has a direct grant or another role with `EXECUTE` | that person reads in clear | read query 1, name them |\n| 3. a user audience for people who read through roles | they read in clear | a role audience, or `ALL` |\n| 4. tagged only the top view | the views below read in clear for whoever may read them | tag the lowest view; grant views, not the database |\n| 5. re-applied a view's file without its `TAGS` | the tag is gone, and the mask | the tags live in the view's file |\n| 6. `DROP ROLE` of a role named in a policy's audience | the policy stops being evaluated for **every** role in it; `DESC VQL` still says `ENABLED = TRUE` and ends with `# Invalid object policy`; query 2 says `INVALID` | read query 2 and the audiences before dropping a role. Afterwards two repairs, both a yes: re-create the role empty — the policy is validated and comes back unchanged — or `CREATE OR REPLACE` the policy without it. Which one depends on why the role went: ask |\n| 7. `DROP TAG … CASCADE` | the tag and every policy that names it are deleted (without `CASCADE`: `Some elements depend on`) | query 3 first |\n| 8. removed a `GRANT` line from a role's file and re-applied it | nothing revoked | `ALTER ROLE … REVOKE` — a yes |\n| 9. left a type out of the masking list | that column comes back `NULL` | name texts, numbers and dates |\n| 10. masked a column people group, filter or join on | groups collapse into `********`, filters find nothing | say it in the message |\n| 11. `ALTER DATABASE … GRANT … TO ROLE` with a misspelled role | `ok`, nothing granted | read query 1 back |\n| 12. trusted `GET_CATALOG_EFFECTIVE_PERMISSIONS` | `rowpermissions = true` for a user who reads in clear | impersonate |\n| 13. a row filter with `VIEWS TAGGED` while the tag is on a column | created, `OK`, filters nothing | `COLUMNS TAGGED` |\n| 14. a per-person filter checked with `impersonate_roles` | no rows — the session's user is `__roles_impersonator_user__` | `impersonate_user`, per person |\n| 15. checked a revoke as the person inside the database itself | `CONNECT` is not checked there: a person with one `EXECUTE` left still read | the impersonated query from another database, `\u003Cdb>.\u003Cview>` |\n| 16. `ALTER USER … REVOKE` of a role or privilege the person does not have | `ok` | `CATALOG_PERMISSIONS()` for the person, after |\n\n## Common mistakes\n\n| You wrote | Server says | Fix |\n|---|---|---|\n| a policy without `ENABLED` | `Syntax error: Exception parsing query near 'AUDIENCE'` | `ENABLED = TRUE` |\n| `MASKING` without `FILTER = ''` | `Syntax error: Exception parsing query near 'MASKING'` | `FILTER = '' MASKING …` |\n| `MASKING ANY ( t )` with no `WITH` | `Syntax error: Cannot invoke \"…getDefaultMasking()\" because \"mt\" is null` | `WITH ( HIDE ) ( texts WITH …, numbers WITH …, datetimes WITH … )` |\n| an audience role that does not exist | `There was an error creating the global security policy. Role 'x' does not exist` | the role first; check the spelling with `--max-rows` |\n| a tag that does not exist | `The following tags do not exist: 'x'` | `CREATE OR REPLACE TAG` first |\n| `CREATE USER u EXTERNAL GRANT ROLE r` | `Syntax error … near 'GRANT'` | accounts are the administrator's; a role is given with `ALTER USER u GRANT ROLE r` |\n| an impersonated query | `This user cannot impersonate. Only users with role 'impersonator' can impersonate.` | the check is unverified — hand the human the queries and the expected answers; ask the administrator |\n| an impersonated query | `The user does not have CONNECT privileges on the database '\u003Cdb>'` | a real answer: that user cannot read the database |\n| `GET_USERS_WITH_ROLE()` with only `role` | `No search methods ready to be run. … include_indirect_roles` | `WHERE role = 'r' AND include_indirect_roles = true` |\n\n## Reference\n\n- `references\u002Fpolicies.md` — the `CREATE GLOBAL_SECURITY_POLICY` grammar, every masking\n  expression as measured, the row filter and deny, enabling and disabling, the audience\n  table per grant path, what removes or disables a policy silently, and its errors.\n- `references\u002Fprivileges.md` — roles, users and grants (`CREATE`\u002F`ALTER ROLE`, `ALTER USER`,\n  `ALTER DATABASE … TO`), the privilege list, column privileges, owners (`CHOWN`), the\n  permission procedures and what each one proves, and impersonation.\n",{"name":5,"description":7},"main","fd41fb386b25141c459264d10a1c49bd3b4f4c4b","skills\u002Fsecurity","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills","https:\u002F\u002Fgithub.com\u002Fkirivandubai","2026-10-29 13:49:22",{"source":8,"skillName":5,"commands":30,"prompt":40,"zipAvailable":41},[31,34,37],{"tool":32,"command":33},"npx","npx skills add kirivandubai\u002Fdenodo_wipe_coading_skills --skill security",{"tool":35,"command":36},"bunx","bunx skills add kirivandubai\u002Fdenodo_wipe_coading_skills --skill security",{"tool":38,"command":39},"pnpm","pnpm dlx skills add kirivandubai\u002Fdenodo_wipe_coading_skills --skill security","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fsecurity\u002FSKILL.md",true,[43,49],{"path":44,"resourceType":45,"resourceTypeDesc":46,"sizeBytes":47,"sourceHtmlUrl":48},"references\u002Fpolicies.md","REFERENCE","参考文档",10771,"https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fsecurity\u002Freferences\u002Fpolicies.md",{"path":50,"resourceType":45,"resourceTypeDesc":46,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fprivileges.md",8921,"https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fsecurity\u002Freferences\u002Fprivileges.md",[54,61,68,75,82,89],{"skillNo":55,"name":56,"nameZh":6,"description":57,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":18},"01M7C2NGAJEKKG90BG9Q1Z2M96","vql","Use when doing anything with Denodo — creating or changing a database, folder, tag, data source, wrapper, base, derived, interface or metric view, association, cache, role or security policy, remote table or summary, a Data Marketplace object, a regression test or a Scheduler job; writing or fixing VQL, or an expression in a SELECT — text, dates, casts; deciding where an object lives and what to name it; \"where do I start\", \"build a mart\", \"connect a source\", \"vibe-code on Denodo\". Start here when the request is ambiguous or names no single object.","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fvql\u002FSKILL.md",[],[],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":18},"01M7C2N5W4203SQ2TRYT6E0VSR","views","Use when building on top of views that already exist in Denodo 9.5 — a derived view (a join, an aggregate, a mart, a filtered or parameterised view, a UNION ALL of views of one entity, FLATTEN of a JSON array into rows, NEST of rows into an array), an interface view, or an association between two views. Also before changing a view others depend on — \"what uses this view\", \"can I drop this column\", \"where does this field come from\" — for whether a mart over a database runs there or pulls every row into Denodo, for \"combine these sources into one view\", \"expose this as a data product\", and for a view created without an error that fails on SELECT or turns up INVALID. Not for sources and base views — \u002Fdenodo:datasources; databases, folders, VDP tags — \u002Fdenodo:catalog; a view's cache — \u002Fdenodo:cache; a copy stored as a table — \u002Fdenodo:materialize; a metric view — \u002Fdenodo:metrics.","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fviews\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":18},"01M7C2MY48XXHR87G39R0F0VTK","testing","Use when a Denodo 9.5 project needs regression tests, or its tests need running or fixing — .denodotest files that the Denodo Testing Tool runs in CI (%EXECUTION, %RESULTS[data|query|csv|exception], configuration.properties, denodo-test.sh) — \"add tests for this mart\", \"a safety net before I rewrite these views\", \"make sure the dashboards see no difference\", \"what should CI check on our data products\", \"prove the mart is right\", \"CI is red on the Denodo tests\", a test that fails on row order, on a column it did not ask for or on a value padded with spaces. Not for checking a new view once, right after creating it — \u002Fdenodo:views.","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Ftesting\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":18},"01M7C2MMFX40K0WTDNW05VCQEJ","semantics","Use when the views of an existing Denodo 9.5 database have to be understood by people and by AI consumers — the Denodo MCP Server, Assisted Query in the Data Marketplace, the AI SDK. Auditing a database for views and fields without descriptions, primary keys that are missing or not unique, missing associations and the VDP tag the MCP Server shows views by; writing view, field, association and tag descriptions from the data, and primary keys. Also for \"describe \u002F document these views\", \"prepare this database for AI, MCP or Assisted Query\", \"why does the agent not see this view through MCP\", a description that says the wrong thing. Not for building or changing what a view computes — \u002Fdenodo:views; not for marketplace-only metadata (logical names, property groups, marketplace tags) — \u002Fdenodo:marketplace.","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fsemantics\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":18},"01M7C2KYWB8B6ZC026CT3Z1MZA","scheduler","Use when something in Denodo 9.5 has to happen on a schedule, or a scheduled run misbehaves — refreshing the cache of a view every night or every hour, running REFRESH of a remote table or a procedure at a set time, exporting a view to a CSV file every week or month, running a Denodo Scheduler job now, stopping, pausing, enabling or deleting it, reading what last night's run did or why it failed, \"the cache doubles after every refresh\", \"the job fires at the wrong hour\", \"the export file is missing or has no header\". Denodo Scheduler projects, jobs, cron triggers and reports, through the REST API of the Scheduler administration tool. Not for switching a cache on or loading it once (\u002Fdenodo:cache), not for creating the remote table or summary itself (\u002Fdenodo:materialize).","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fscheduler\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":18},"01M7C2KQBX7XS4JYW68BBRPXQN","procedures","Use when a Denodo 9.5 stored procedure is in play — calling one the server already ships (SELECT … FROM GET_VIEWS() \u002F USED_BY() … WHERE input_… = …, the CALL statement, finding which predefined procedure answers a question and what parameters it takes), writing your own procedural logic in VQL (CREATE OR REPLACE VQL PROCEDURE … BEGIN … END, local variables, IF\u002FLOOP\u002FCASE, cursors, exceptions, EXECUTE of DDL), or importing a Java one from a JAR (CREATE PROCEDURE … CLASSNAME … JARS). Also for \"run the stats procedure\", \"loop over rows and create views\", \"call that procedure from a view\". Not for introspecting a JDBC source into base views — that is \u002Fdenodo:datasources.","https:\u002F\u002Fgithub.com\u002Fkirivandubai\u002Fdenodo_wipe_coading_skills\u002Fblob\u002Fmain\u002Fskills\u002Fprocedures\u002FSKILL.md",[],[]]