diff --git a/CHANGELOG.adoc b/CHANGELOG.adoc
index 40af30364bcb198bca27592f0f2aa5f0745b5e09..df3107604a486b9411e2fea4ae6f0616d3635964 100644
--- a/CHANGELOG.adoc
+++ b/CHANGELOG.adoc
@@ -4,6 +4,17 @@
 This is a summary of all notable changes to the Antora Collector Extension by release.
 For a detailed view of what's changed, refer to the {url-repo}/commits[commit history] of this project.
 
+== Unreleased
+
+=== Added
+
+* add an optional `clean` configuration key on each collector entry to specify a directory to clean (i.e., remove) (#10)
+
+=== Changed
+
+* don't remove the scan dir unless the `clean: true` is specified on the scan entry (#10)
+* process cleans per collector entry rather than before running collector on the origin (#10)
+
 == 1.0.0-alpha.3 (2022-11-12)
 
 === Changed
diff --git a/docs/modules/ROOT/pages/configuration-keys.adoc b/docs/modules/ROOT/pages/configuration-keys.adoc
index fbebf45df8a7369e099cb422b697a4b13c128103..05930740a98e60a15b62c3ef7a01b84a985d61ab 100644
--- a/docs/modules/ROOT/pages/configuration-keys.adoc
+++ b/docs/modules/ROOT/pages/configuration-keys.adoc
@@ -6,8 +6,12 @@ The `ext` key is the designated area in the component version descriptor for ext
 [#collector-key]
 == collector key
 
-The external commands the extension runs and scan settings are defined under the `collector` key nested under the `ext` key in an [.path]_antora.yml_ file.
-These commands and settings are applied to a git reference when Antora visits the content roots of the content sources defined in the playbook to produce a content aggregate.
+The `collector` key in the component version descriptor ([.path]_antora.yml_) defines the configuration for the clean, run, and scan steps that are carried out on the origin (a git reference) in which the component version descriptor is found.
+All three steps are optional.
+The settings for these steps are defined using keys of the same names under the `collector` key.
+The extension runs the steps in aforementioned order, so it's typical to specify the keys in this order as well.
+
+Here's an example of a component version descriptor that configures the collector extension.
 
 .antora.yml
 [,yaml]
@@ -19,7 +23,9 @@ nav:
 - modules/ROOT/nav.adoc
 ext:
   collector: # <.>
-  - run: # <.>
+  - clean: # <.>
+      dir: build/generated # <.>
+    run: # <.>
       command: ./gradlew --console rich generateContent # <.>
     scan: # <.>
     - dir: build/generated # <.>
@@ -28,13 +34,55 @@ ext:
       dir: artifacts
 ----
 <.> `collector` key
+<.> `clean` key
+<.> `dir` key that specifies the directory to clean
 <.> `run` key
-<.> `command` key
+<.> `command` key that specifies the command to run
 <.> `scan` key
-<.> `dir` key
-<.> `files` key
+<.> `dir` key that specifies the directory to scan
+<.> `files` key that provides a pattern of files to scan
+
+The `collector` key accepts an array (i.e., list) of collector entries.
+If there's only a single entry, the array can be replaced by a map for a single entry (i.e., no preceding hyphen).
+
+If the `collector` key isn't set in a component version descriptor, or its value is falsy, the extension won't run on that git reference.
+
+[#clean-key]
+== clean key
 
-If `collector` isn't set in a component version descriptor, or its value is falsy, the Collector extension won't run on that git reference.
+The `clean` key must be nested under the `collector` key.
+The `clean` key accepts a list of built-in key-value pairs that configure the directory (`dir`) to clean.
+
+.antora.yml
+[,yaml]
+----
+name: colorado
+title: Colorado
+version: '5.6.0'
+ext:
+  collector:
+  - clean:
+      dir: build/generated
+    run:
+      command: ./gradlew --console rich generateContent
+    scan:
+    - dir: build/generated
+      files: '**/*.adoc'
+    - dir: build/log
+  - scan:
+      dir: artifacts
+----
+
+The `clean` key accepts an array of items (i.e., list) that each have a `dir` key.
+Each `clean` entry is invoked sequentially in the order specified in the array.
+If there's only a single entry, the array can be replaced by a map for a single entry.
+
+If the directory to clean is the same as the directory to scan, the clean entry can be created implicitly by setting the `clean` key on the scan entry to `true` (e.g., `clean: true`).
+See <<scan-key>>.
+
+Note that when running collector on a content source without a worktree (such as a remote repository), the worktree will always start out in a clean state.
+When the content source is local and has a worktree, the worktree may already contain files that are untracked or ignored by git.
+In this case, the first clean entry can be important to ensuring a predictable result.
 
 [#run-key]
 == run key
@@ -54,6 +102,7 @@ ext:
       command: ./gradlew --console rich generateContent
     scan:
       dir: build/generated
+      clean: true
   - run:
       command: stats --validate package
     scan:
@@ -66,7 +115,7 @@ Each `run` key is invoked sequentially in the order specified in the array.
 [#scan-key]
 == scan key
 
-The `scan` key can be nested under the `collector` key or under a `run` key.
+The `scan` key can be nested under the `collector` key if there's only a single collector entry, or in an item in the array of entries on the `collector` key.
 The `scan` key accepts a list of built-in key-value pairs that configure the scan directory (`dir`) and file filter (`files`).
 
 .antora.yml
@@ -82,20 +131,18 @@ ext:
     scan:
     - dir: build/generated
       files: '**/*.adoc'
+      clean: true
     - dir: build/log
+      clean: true
   - scan:
       dir: artifacts
 ----
 
-The `scan` key can be specified multiple times.
-Each `scan` key is invoked sequentially in the order specified in the array.
-When a `scan` key is set directly under `collector` (not nested under `run`), it isn't associated with the outcome of an external command.
+The `scan` key accepts an array of items (i.e., list) that have the nested keys.
+Each `scan` entry is invoked sequentially in the order specified in the array.
+If there's only a single entry, the array can be replaced by a map of a single entry.
 
-CAUTION: The current assumption is that the scan directories are generated by a run command.
-When using a local content source, the scan directories are removed before any commands are run.
-That means that if you point a scan at a directory that already exists, it will be removed.
-This behavior is under review and will likely be changed or made configurable in the future.
-As a workaround, a run command can be used to copy files from a permanent directory to a temporary one for scanning.
+If the `clean: true` key is set on the entry, that implicitly creates a clean entry with the same dir.
 
 [#collector-reference]
 == Available Collector keys
@@ -106,11 +153,19 @@ The table below lists the keys that can be defined in a component version descri
 |===
 |Collector Keys |Description |Default |Values
 
+|`clean.dir`
+|Defines the location of a directory to clean before processing the rest of the collector entry (i.e., before run and scan).
+The value of the `dir` key is a string path for a single directory.
+If the value is `.` or starts with `./`, it's resolved starting from the start path.
+Otherwise, the path is resolved relative to the worktree.
+|Not set
+|Path relative to the content root
+
 |`run.command`
 |Specifies a command to run for the current Antora content root.
 The command is run from the root of the worktree for the git reference unless the `dir` key is specified.
 The `dir` key specifies an alternate working directory (e.g., `scripts`).
-If the value is `.` or starts with `./`, the value is resolved from the start path.
+If the value is `.` or starts with `./`, the value is resolved starting from the start path.
 The commands are invoked sequentially in the order specified in this array.
 If the value of `collector` is `false`, no commands will be invoked.
 |Not set
@@ -118,18 +173,17 @@ If the value of `collector` is `false`, no commands will be invoked.
 
 |`run.dir`
 |Defines the location (pwd) from where the command is run.
-The value of the `dir` key is a string.
-If the value is `.` or starts with `./`, it's resolved from the start path.
+The value of the `dir` key is a string path for a single directory.
+If the value is `.` or starts with `./`, it's resolved starting from the start path.
 Otherwise, the path is resolved relative to the worktree.
 |Not set
 |Path relative to the content root
 
 |`scan.dir`
 |Defines the location from where the extension collects the generated files after the previous command is complete.
-The extension then imports the collected files into the bucket in the content aggregate .
-The value of the `dir` key is an array.
-Each entry in that array provides a path for a single directory.
-If the value is `.` or starts with `./`, it's resolved from the start path.
+The extension then imports the collected files into the bucket in the content aggregate.
+The value of the `dir` key is a string path for a single directory.
+If the value is `.` or starts with `./`, it's resolved starting from the start path.
 Otherwise, the path is resolved relative to the worktree.
 |Not set
 |Path relative to the content root
diff --git a/packages/collector-extension/lib/index.js b/packages/collector-extension/lib/index.js
index fb3f0bbb6c75c77578e8c36773e18b5b6be355be..2db31eff6f6560cf04f23638afe836748a265834 100644
--- a/packages/collector-extension/lib/index.js
+++ b/packages/collector-extension/lib/index.js
@@ -33,10 +33,14 @@ module.exports.register = function () {
         if (Array.isArray(collectorConfig) && !collectorConfig.length) continue
         const worktreeDir = worktree || ospath.join(cacheDir, generateWorktreeFolderName({ url, gitdir, worktree }))
         const expandPathContext = { base: worktreeDir, cwd: worktreeDir, dot: ospath.join(worktreeDir, startPath) }
-        const scanDirs = new Set()
         const collectors = (Array.isArray(collectorConfig) ? collectorConfig : [collectorConfig]).map((collector) => {
-          const { run: runConfig = {}, scan: scanConfig = [] } = collector
+          const { clean: cleanConfig = [], run: runConfig = {}, scan: scanConfig = [] } = collector
+          let cleans
           return {
+            clean: (cleans = (Array.isArray(cleanConfig) ? cleanConfig : [cleanConfig]).reduce((accum, clean) => {
+              if (typeof clean.dir === 'string') accum.push({ dir: expandPath(clean.dir, expandPathContext) })
+              return accum
+            }, [])),
             run: {
               ...runConfig,
               cwd: typeof runConfig.dir === 'string' ? expandPath(runConfig.dir, expandPathContext) : worktreeDir,
@@ -44,21 +48,22 @@ module.exports.register = function () {
             scan: (Array.isArray(scanConfig) ? scanConfig : [scanConfig]).reduce((accum, scan) => {
               if (typeof scan.dir === 'string') {
                 const dir = expandPath(scan.dir, expandPathContext)
-                scanDirs.add(dir)
+                if (scan.clean) cleans.push({ dir })
                 accum.push({ ...scan, dir })
               }
               return accum
             }, []),
           }
         })
-        if (worktree) {
-          for (const scanDir of scanDirs) await fsp.rm(scanDir, { recursive: true, force: true })
-        } else {
+        if (!worktree) {
           const cache = gitCache[gitdir] || (gitCache[gitdir] = {})
           const ref = `refs/${reftype === 'branch' ? 'head' : reftype}s/${refname}`
           await prepareWorktree({ fs, cache, dir: worktreeDir, gitdir, ref, remote, bare: worktree === undefined })
         }
-        for (const { run, scan: scans } of collectors) {
+        for (const { clean: cleans, run, scan: scans } of collectors) {
+          for (const clean of cleans) {
+            await fsp.rm(clean.dir, { recursive: true, force: true })
+          }
           const { cwd, command, local } = run
           if (command) {
             let cmd = command
diff --git a/packages/collector-extension/test/collector-extension-test.js b/packages/collector-extension/test/collector-extension-test.js
index 295da436bfb3a5a1281855582c46fc1ca3b8ee5a..e442624f1d42aa682ca54268e1da5b494454648d 100644
--- a/packages/collector-extension/test/collector-extension-test.js
+++ b/packages/collector-extension/test/collector-extension-test.js
@@ -247,6 +247,7 @@ describe('collector extension', () => {
 
     it('should populate properties of file collected from concrete worktree', async () => {
       const collectorConfig = {
+        clean: { dir: 'build' },
         run: { command: 'node .gen-start-page.js' },
         scan: { dir: 'build' },
       }
@@ -513,6 +514,26 @@ describe('collector extension', () => {
       })
     })
 
+    it('should not delete files in scan dir of worktree before running scan', async () => {
+      const collectorConfig = {
+        scan: { dir: 'other-docs' },
+      }
+      await runScenario({
+        repoName: 'test-at-start-path',
+        startPath: 'docs',
+        collectorConfig,
+        local: true,
+        before: (contentAggregate) => {
+          expect(contentAggregate).to.have.lengthOf(1)
+          expect(contentAggregate[0].files).to.be.empty()
+        },
+        after: (contentAggregate) => {
+          expect(contentAggregate[0].files).to.have.lengthOf(1)
+          expect(contentAggregate[0].files[0].path).to.equal('modules/ROOT/pages/outside-start-path.adoc')
+        },
+      })
+    })
+
     it('should rebase files if base key is set on scan', async () => {
       const collectorConfig = {
         scan: { dir: 'code', base: 'modules/extend/examples' },
@@ -651,6 +672,44 @@ describe('collector extension', () => {
       })
     })
 
+    it('should clean specified dirs per collector entry', async () => {
+      const collectorConfig = [
+        {
+          run: { command: 'node .gen-files.js' },
+        },
+        {
+          run: { command: 'node .gen-component-desc.js' },
+        },
+        {
+          clean: [{ dir: 'build/modules/ROOT/examples' }, { dir: 'build/modules/ROOT/partials' }],
+          run: { command: 'node .gen-start-page.js' },
+          scan: [
+            { dir: 'build', files: 'antora.yml' },
+            { dir: 'build', files: 'modules/**/*' },
+          ],
+        },
+      ]
+      await runScenario({
+        repoName: 'test-at-root',
+        collectorConfig,
+        before: (contentAggregate) => {
+          expect(contentAggregate).to.have.lengthOf(1)
+          expect(contentAggregate[0].version).to.eql('main')
+          expect(contentAggregate[0].files).to.be.empty()
+        },
+        after: (contentAggregate) => {
+          expect(contentAggregate).to.have.lengthOf(1)
+          const bucket = contentAggregate[0]
+          expect(bucket.version).to.equal('1.0.0')
+          expect(bucket.files).to.have.lengthOf(1)
+          expect(bucket.files.map((it) => it.path)).to.have.members(['modules/ROOT/pages/index.adoc'])
+          expect(
+            contentAggregate[0].files.find((it) => it.path === 'modules/ROOT/pages/index.adoc').contents.toString()
+          ).to.include('= Start Page')
+        },
+      })
+    })
+
     it('should only collect files that match the scan pattern', async () => {
       const collectorConfig = {
         run: { command: 'node .gen-files.js' },
@@ -720,7 +779,7 @@ describe('collector extension', () => {
       expect(getCollectorCacheDir()).to.not.be.a.path()
     })
 
-    it('should reuse worktree if available', async () => {
+    it('should reuse worktree without cleaning if no clean step is specified', async () => {
       const collectorConfig = {
         run: { command: 'node .gen-start-page.js' },
         scan: { dir: 'build' },
@@ -729,23 +788,28 @@ describe('collector extension', () => {
         repoName: 'test-at-root',
         local: true,
         collectorConfig,
-        before: (contentAggregate) => {
+        before: async (contentAggregate) => {
           expect(contentAggregate).to.have.lengthOf(1)
           expect(contentAggregate[0].files).to.be.empty()
+          const worktree = contentAggregate[0].origins[0].worktree
+          const untrackedPagePath = ospath.join(worktree, 'build/modules/ROOT/pages/untracked.adoc')
+          await fsp.mkdir(ospath.dirname(untrackedPagePath), { recursive: true })
+          await fsp.writeFile(untrackedPagePath, '= Untracked', 'utf8')
         },
         after: (contentAggregate) => {
-          expect(contentAggregate[0].files).to.have.lengthOf(1)
+          expect(contentAggregate[0].files).to.have.lengthOf(2)
           const worktree = contentAggregate[0].origins[0].worktree
           expect(ospath.join(worktree, 'build')).to.be.a.directory()
           expect(ospath.join(worktree, 'build/modules/ROOT/pages/index.adoc')).to.be.a.file()
+          expect(ospath.join(worktree, 'build/modules/ROOT/pages/untracked.adoc')).to.be.a.file()
         },
       })
     })
 
-    it('should clean scan dir in worktree before running command(s)', async () => {
+    it('should clean scan dir in worktree before running command(s) if clean key on scan entry is true', async () => {
       const collectorConfig = {
         run: { command: 'node .gen-start-page.js' },
-        scan: { dir: 'build' },
+        scan: { clean: true, dir: 'build' },
       }
       await runScenario({
         repoName: 'test-at-root',
@@ -767,6 +831,7 @@ describe('collector extension', () => {
 
     it('should run specified command in temporary worktree if repository is local and reference is not worktree', async () => {
       const collectorConfig = {
+        clean: { dir: 'build' },
         run: { command: 'node .gen-start-page.js' },
         scan: { dir: 'build' },
       }
@@ -789,6 +854,7 @@ describe('collector extension', () => {
 
     it('should run specified command in each branch that defines collector config', async () => {
       const collectorConfig = {
+        clean: { dir: 'build' },
         run: { command: 'node .gen-start-page.js' },
         scan: { dir: 'build' },
       }
@@ -894,6 +960,7 @@ describe('collector extension', () => {
 
     it('should not modify index of local repository when checking out ref to worktree', async () => {
       const collectorConfig = {
+        clean: { dir: 'build' },
         run: { command: 'node .gen-start-page.js' },
         scan: { dir: 'build' },
       }