A curated collection of Agent Skills for working with dbt, to help AI agents understand and execute dbt workflows more effectively.
70
88%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
View guide
Low
Low-risk findings worth noting
{
"from_version": "1.8",
"warehouse": "spark",
"target_version": "1.12",
"count": 16,
"issues": [
{
"issue_id": "1_8_001",
"sort_order": 6001,
"change": "on-run-start hook failure now skips all downstream nodes instead of continuing",
"action": "Set flags.skip_nodes_if_on_run_start_fails: false to keep the old behavior, or make on-run-start hooks reliable.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "A failing on-run-start hook aborts the whole run's nodes rather than letting them execute. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.8",
"to_version": "1.9",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "skip_nodes_if_on_run_start_fails",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project defines `on-run-start` hooks.\n- Check dbt_project.yml for a non-empty `on-run-start:` list.\n- No on-run-start hooks -> the gated behavior cannot occur -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_8_001`. No project code changes.\nOnly pin it when on-run-start hooks exist."
},
"_path": "kb/core/1_8_001.yaml"
},
{
"issue_id": "1_8_002",
"sort_order": 6002,
"change": "state:modified compares more unrendered config values, changing which nodes are selected",
"action": "Set flags.state_modified_compare_more_unrendered_values: false to keep the old comparison.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Slim CI / state:modified selection can pick a different node set, changing what gets built. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.8",
"to_version": "1.9",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "state_modified_compare_more_unrendered_values",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project relies on `state:modified` selection.\n- Look for state comparison in-repo: a `selectors.yml` using `state:modified`,\n or CI config / job commands in the repo passing `--select state:modified`.\n- If nothing in the repo uses state:modified, this cannot be confirmed from the\n repo alone -> skipped-not-present, and mention it in the report so the user can\n decide (their CI may use it out of repo).",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_8_002`. No project code changes.\nOnly pin it when the repo shows state:modified usage."
},
"_path": "kb/core/1_8_002.yaml"
},
{
"issue_id": "1_8_003",
"sort_order": 6003,
"change": "custom microbatch strategies must use batched execution",
"action": "Set flags.require_batched_execution_for_custom_microbatch_strategy: false to keep the old behavior, or adopt batched execution.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "A custom microbatch incremental strategy that is not batch-aware behaves differently or errors. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.8",
"to_version": "1.9",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_batched_execution_for_custom_microbatch_strategy",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project defines a CUSTOM microbatch incremental strategy.\n- Check macros/ for a macro named `get_incremental_microbatch_sql` or a\n materialization/strategy override handling `microbatch`, and models configured\n with `incremental_strategy: microbatch`.\n- Built-in microbatch usage alone is NOT affected -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_8_003`. No project code changes.\nOnly pin it when a custom microbatch strategy macro exists."
},
"_path": "kb/core/1_8_003.yaml"
},
{
"issue_id": "1_8_004",
"sort_order": 6004,
"change": "cumulative metric type params must be nested under type_params",
"action": "Set flags.require_nested_cumulative_type_params: false to keep accepting the flat shape, or nest the params.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Semantic-layer cumulative metrics using the flat param shape fail validation. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.8",
"to_version": "1.9",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_nested_cumulative_type_params",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project defines cumulative metrics using the FLAT param\n shape (window/grain_to_date directly under the metric rather than nested under\n `type_params`).\n- Grep metric YAML for `type: cumulative`, then check whether its params are\n nested under `type_params:`.\n- Already nested, or no cumulative metrics -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_8_004`. No project code changes.\nOnly pin it when a flat-shaped cumulative metric exists."
},
"_path": "kb/core/1_8_004.yaml"
},
{
"issue_id": "1_9_001",
"sort_order": 7001,
"change": "macro arguments are validated against their declared signatures",
"action": "Set flags.validate_macro_args: false to keep validation off, or correct the macro argument declarations.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Macros whose calls do not match their documented arguments now warn or error at parse time. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.9",
"to_version": "1.10",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "validate_macro_args",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project has macros whose call sites disagree with their\n declared arguments (the thing validation would now reject).\n- Check macros/ for documented `arguments:` in YAML that do not match the macro\n signature, or calls passing unexpected/missing args.\n- The reliable check: run the parse gate; if it reports a macro-argument\n validation error/warning, the behavior is exhibited.\n- Parse clean with the flag unset -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_9_001`. No project code changes.\nPrefer pinning only when the parse gate actually flags a macro-argument problem."
},
"_path": "kb/core/1_9_001.yaml"
},
{
"issue_id": "1_9_002",
"sort_order": 7002,
"change": "every warning must be handled when warn_error / warn_error_options is set",
"action": "Set flags.require_all_warnings_handled_by_warn_error: false to keep the old partial handling.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Projects using warn_error_options can start failing on warnings that were previously not escalated. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.9",
"to_version": "1.10",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_all_warnings_handled_by_warn_error",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project sets `warn_error` or `warn_error_options`\n (dbt_project.yml `flags:`, profiles.yml, or CI job commands in the repo).\n- Neither set -> the stricter handling changes nothing -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_9_002`. No project code changes.\nOnly pin it when warn_error / warn_error_options is actually configured."
},
"_path": "kb/core/1_9_002.yaml"
},
{
"issue_id": "1_9_003",
"sort_order": 7003,
"change": "generic test arguments must be declared under an `arguments:` property",
"action": "Set flags.require_generic_test_arguments_property: false to keep accepting the old inline shape, or move test args under `arguments:`.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Generic tests passing args inline (not under `arguments:`) fail validation. This flag defaults to True in 1.12, so leaving it unset silently adopts the new behavior.",
"from_version": "1.9",
"to_version": "1.10",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_generic_test_arguments_property",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if generic tests pass arguments inline instead of under an\n `arguments:` property.\n- Grep models/**/*.yml (and any tests/ YAML) for generic tests with sibling keys\n next to the test name that are not `arguments:` (e.g. `accepted_values:` with\n `values:` inline).\n- All generic test args already under `arguments:`, or no parameterized generic\n tests -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_9_003`. No project code changes.\nOnly pin it when inline generic-test arguments are present."
},
"_path": "kb/core/1_9_003.yaml"
},
{
"issue_id": "1_10_001",
"sort_order": 8001,
"change": "resource names must be unique within a project",
"action": "Set flags.require_unique_project_resource_names: false to keep allowing duplicates, or rename the colliding resources.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Projects with two resources sharing a name across paths start erroring at parse time. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.10",
"to_version": "1.11",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_unique_project_resource_names",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project has DUPLICATE resource names within itself.\n- List model/seed/snapshot/test file stems and look for the same name declared\n more than once across different paths.\n- All names unique -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_10_001`. No project code changes.\nOnly pin it when an actual duplicate resource name exists."
},
"_path": "kb/core/1_10_001.yaml"
},
{
"issue_id": "1_10_002",
"sort_order": 8002,
"change": "ref() resolves against the node's own package before the root project",
"action": "Set flags.require_ref_searches_node_package_before_root: false to keep root-first resolution.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "In projects with packages that shadow root model names, ref() can resolve to a different node. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.10",
"to_version": "1.11",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_ref_searches_node_package_before_root",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if an installed package defines a model name that ALSO exists in\n the root project (the shadowing case where resolution order changes).\n- Check packages.yml/dependencies.yml for packages, then compare package model\n names against root model names (dbt_packages/ if installed).\n- No packages, or no name collision -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_10_002`. No project code changes.\nOnly pin it when a package/root model name collision exists."
},
"_path": "kb/core/1_10_002.yaml"
},
{
"issue_id": "1_11_001",
"sort_order": 9001,
"change": "generate_schema_name must return a valid, non-empty schema name",
"action": "Set flags.require_valid_schema_from_generate_schema_name: false to keep the lax behavior, or fix the macro.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "A generate_schema_name override returning None/empty now errors instead of silently falling back. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_valid_schema_from_generate_schema_name",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project overrides `generate_schema_name` (or\n `generate_schema_name_for_env`) in macros/.\n- Inspect the override for a path that can return None/empty (e.g. a missing\n `{% else %}` branch, or returning `custom_schema_name` unguarded).\n- No override, or the override always returns a non-empty schema ->\n skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_11_001`. No project code changes.\nOnly pin it when a generate_schema_name override can return empty."
},
"_path": "kb/core/1_11_001.yaml"
},
{
"issue_id": "1_11_002",
"sort_order": 9002,
"change": "sql_header must be declared in test configs to take effect",
"action": "Set flags.require_sql_header_in_test_configs: false to keep the old handling.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Tests relying on an implicitly inherited sql_header may lose it. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_sql_header_in_test_configs",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if tests rely on a `sql_header` they do not declare themselves\n (i.e. inherited from a model/project-level config).\n- Grep for `sql_header` in dbt_project.yml, model configs, and test configs; the\n issue is present when a test needs it but only a non-test config sets it.\n- No sql_header anywhere, or tests declare their own -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_11_002`. No project code changes.\nOnly pin it when a test depends on an inherited sql_header."
},
"_path": "kb/core/1_11_002.yaml"
},
{
"issue_id": "1_11_003",
"sort_order": 9003,
"change": "analyses use corrected fully-qualified names",
"action": "Set flags.require_corrected_analysis_fqns: false to keep the previous FQNs.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Selection and documentation paths for analyses change, affecting --select on analyses. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_corrected_analysis_fqns",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project has an analyses/ directory with content AND\n something selects analyses by FQN (selectors.yml or in-repo job commands).\n- Empty/absent analyses/ -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_11_003`. No project code changes.\nOnly pin it when analyses exist and are selected by FQN."
},
"_path": "kb/core/1_11_003.yaml"
},
{
"issue_id": "1_11_004",
"sort_order": 9004,
"change": "source and semantic model names may not contain spaces",
"action": "Set flags.require_source_and_semantic_model_names_without_spaces: false to keep allowing spaces, or rename them.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Sources/semantic models with spaces in their names start erroring rather than warning. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "require_source_and_semantic_model_names_without_spaces",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if a SOURCE or SEMANTIC MODEL name contains a space.\n- Grep source/semantic model YAML for `- name:` values containing a space.\n (Model names with spaces are a different, pre-1.8 issue \u2014 not this one.)\n- No spaces in source/semantic model names -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_11_004`. No project code changes.\nOnly pin it when a source/semantic model name actually contains a space."
},
"_path": "kb/core/1_11_004.yaml"
},
{
"issue_id": "1_11_005",
"sort_order": 9005,
"change": "Jinja-suffixed file extensions (e.g. .sql.jinja) are recognized",
"action": "Leave flags.allow_jinja_file_extensions: false to keep the previous file discovery.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Enabling it changes which files dbt picks up as models; leaving it false preserves current discovery. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "allow_jinja_file_extensions",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- This flag ENABLES new file-extension discovery; it defaults to false, so the\n project's current behavior is already the legacy one.\n- Only relevant if the project has files with a `.jinja`-style suffix it expects\n dbt to pick up.\n- No such files -> skipped-not-present (pinning would be pure noise).",
"fixing": "Normally skipped-not-present: false is already the default and the legacy\nbehavior. Pin via `tools.py set-flag --issue-id 1_11_005` only if the project has\nJinja-suffixed files and you need to lock the current discovery behavior."
},
"_path": "kb/core/1_11_005.yaml"
},
{
"issue_id": "1_11_006",
"sort_order": 9006,
"change": "versioned models get a `latest` version pointer by default",
"action": "Leave flags.latest_version_pointer_enabled_by_default: false to keep explicit-only version pointers.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "For versioned models, ref() without a version can resolve differently. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "latest_version_pointer_enabled_by_default",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project uses MODEL VERSIONS (`versions:` in model YAML)\n and has `ref()` calls to a versioned model without an explicit version.\n- No versioned models -> the pointer default cannot matter -> skipped-not-present.",
"fixing": "Pin the flag via `tools.py set-flag --issue-id 1_11_006`. No project code changes.\nOnly pin it when versioned models exist and are ref'd without a version."
},
"_path": "kb/core/1_11_006.yaml"
},
{
"issue_id": "1_11_007",
"sort_order": 9007,
"change": "v2 catalog integration replaces the previous catalog behavior (all adapters)",
"action": "Leave flags.use_catalogs_v2: false to keep the previous catalog behavior.",
"category": "Behavior",
"component": "core",
"adapter_type": null,
"impact": "Catalog/metadata resolution changes across adapters; leaving it false preserves current behavior. This flag still defaults to False in 1.12; pinning it makes that explicit and survives a future flip.",
"from_version": "1.11",
"to_version": "1.12",
"automation_type": "behavior_flag",
"behavior_flag": {
"name": "use_catalogs_v2",
"set_to": false
},
"out_of_repo_risk": false,
"environment_change": false,
"context": {
"detection": "- Applies only if the project defines catalog configuration (e.g. a `catalogs:`\n block or catalog-related model configs) whose resolution v2 would change.\n- Defaults to false, so a project with no catalog config already has the legacy\n behavior -> skipped-not-present (pinning would be pure noise).",
"fixing": "Normally skipped-not-present: false is already the default and the legacy\nbehavior. Pin via `tools.py set-flag --issue-id 1_11_007` only if the project has\ncatalog configuration whose behavior you need to lock."
},
"_path": "kb/core/1_11_007.yaml"
}
]
}.changes
.claude
skills
auditing-skills
.claude-plugin
.cursor-plugin
.github
ISSUE_TEMPLATE
scripts
skills
dbt
.claude-plugin
.cursor-plugin
skills
adding-dbt-unit-test
references
answering-natural-language-questions-with-dbt
building-dbt-semantic-layer
configuring-dbt-mcp-server
fetching-dbt-docs
scripts
maintaining-dbt-documentation
running-dbt-commands
troubleshooting-dbt-job-errors
references
using-dbt-for-analytics-engineering
using-dbt-state
working-with-dbt-mesh
dbt-extras
.claude-plugin
skills
creating-mermaid-dbt-dag
dbt-migration
.claude-plugin
skills
migrating-dbt-core-to-v2
migrating-dbt-project-across-platforms
upgrading-dbt
references
scripts