@@ -40,6 +40,17 @@ Actions with special purposes and unlikely to be used directly:
 
 All advanced setup code scanning workflows must have the `security-events: write` permission. Workflows in private repositories must additionally have the `contents: read` permission. For more information, see "[Assigning permissions to jobs](https://docs.github.com/en/actions/using-jobs/assigning-permissions-to-jobs)."
 
+### Skipping analysis during pull request iteration
+
+The `init` action supports two opt-in inputs, both disabled by default:
+
+- `skip-if-draft: true` skips analysis when the pull request event confirms that the PR is a draft. Include `ready_for_review` in the workflow's `pull_request.types` so analysis runs when the PR becomes ready. Confirmed draft skips return before repository configuration lookups, status reporting, tool download, or database initialization.
+- `skip-if-no-language-changes: true` skips a single explicitly selected built-in language only when a complete comparison of the PR event's immutable commit SHAs contains exclusively recognized non-code paths. Source files, workflows, build and dependency configuration, shell scripts, unknown paths, incomplete comparisons, and API errors retain full analysis. This is a conservative path heuristic; leave it disabled when documentation or other allowed non-code files are inputs to code generation.
+
+For example, include `types: [opened, synchronize, reopened, ready_for_review]` under `on.pull_request`, assign an `id` to the `init` step, and enable the desired inputs. The `analysis-skipped` and `analysis-skip-reason` outputs describe the decision. Guard custom build steps with `if: steps.init.outputs.analysis-skipped != 'true'`. The `autobuild`, `analyze`, and post actions handle intentional skips automatically. When analysis runs, it still uses the complete language database rather than analyzing only changed files.
+
+GitHub-managed workflows require integration by their workflow generator to provide the inputs and draft metadata. These inputs do not configure managed Code Quality from a repository workflow file.
+
 ### Build Modes
 
 The CodeQL Action supports different build modes for analyzing the source code. The available build modes are:

@@ -93,6 +93,10 @@ outputs:
     description: Absolute, local path to the directory containing the generated SARIF file.
   sarif-id:
     description: The ID of the uploaded SARIF file.
+  analysis-skipped:
+    description: Whether analysis was intentionally skipped before CodeQL finalization
+  analysis-skip-reason:
+    description: Why analysis was intentionally skipped, if applicable
 runs:
   using: node24
   main: "../lib/analyze-entry.js"

@@ -29,6 +29,24 @@ inputs:
       For more information, see
       https://docs.github.com/en/code-security/code-scanning/creating-an-advanced-setup-for-code-scanning/customizing-your-advanced-setup-for-code-scanning#changing-the-languages-that-are-analyzed.
     required: false
+  skip-if-draft:
+    description: >-
+      When true and the current event confirms this is a draft pull request, skip CodeQL before
+      downloading tools or initializing a database. The workflow must run again on `ready_for_review`
+      to analyze the PR once it is marked ready. For a `dynamic` workflow, the workflow generator can
+      set `CODE_SCANNING_IS_DRAFT=true` to trigger the same early skip. If draft state is unavailable,
+      analysis continues.
+    required: false
+    default: 'false'
+  skip-if-no-language-changes:
+    description: >-
+      When true for a single explicitly requested built-in language, skip CodeQL before tool download
+      and database initialization if the complete pull-request diff contains no relevant source or
+      language configuration files. Unknown, incomplete, or unrecognized diffs fail open and run the
+      full analysis. This uses the complete language database when relevant changes exist; it does
+      not restrict analysis to changed files.
+    required: false
+    default: 'false'
   build-mode:
     description: >-
       The build mode that will be used to analyze the language. This input is only available when
@@ -176,6 +194,10 @@ outputs:
     description: The path of the CodeQL binary used for analysis
   codeql-version:
     description: The version of the CodeQL binary used for analysis
+  analysis-skipped:
+    description: Whether analysis was intentionally skipped before CodeQL initialization
+  analysis-skip-reason:
+    description: Why analysis was intentionally skipped, if applicable
 runs:
   using: node24
   main: '../lib/init-entry.js'