refactor: migrate datastore from data/ to .data/ hidden directory

.claude/skills/swamp-model/SKILL.md

@@ -12,19 +12,19 @@ machine-readable output.
 
 Swamp uses a dual-layer architecture:
 
-- **Data directory (`/data/`)** - Internal storage organized by entity type
+- **Data directory (`/.data/`)** - Internal storage organized by entity type
 - **Logical views (`/models/`)** - Human-friendly symlinked directories
 
 The `/models/` directory provides convenient exploration of each model:
 
 ```
 /models/{model-name}/
-  input.yaml → ../data/inputs/{type}/{id}.yaml
-  resource.yaml → ../data/resources/{type}/{id}.yaml
-  data.yaml → ../data/data/{type}/{id}.yaml
-  outputs/ → ../data/outputs/{type}/{id}/
-  logs/ → ../data/logs/{type}/{id}/
-  files/ → ../data/files/{type}/{id}/
+  input.yaml → ../.data/inputs/{type}/{id}.yaml
+  resource.yaml → ../.data/resources/{type}/{id}.yaml
+  data.yaml → ../.data/data/{type}/{id}.yaml
+  outputs/ → ../.data/outputs/{type}/{id}/
+  logs/ → ../.data/logs/{type}/{id}/
+  files/ → ../.data/files/{type}/{id}/
 ```
 
 This structure is maintained automatically. Use `swamp repo index` to rebuild if

.claude/skills/swamp-workflow/SKILL.md

@@ -12,18 +12,18 @@ machine-readable output.
 
 Swamp uses a dual-layer architecture:
 
-- **Data directory (`/data/`)** - Internal storage organized by entity type
+- **Data directory (`/.data/`)** - Internal storage organized by entity type
 - **Logical views (`/workflows/`)** - Human-friendly symlinked directories
 
 The `/workflows/` directory provides convenient exploration of each workflow:
 
 ```
 /workflows/{workflow-name}/
-  workflow.yaml → ../data/workflows/{id}.yaml
+  workflow.yaml → ../.data/workflows/{id}.yaml
   runs/
     latest → {most-recent-run}/
     {timestamp}/
-      run.yaml → ../data/workflow-runs/{id}/{run-id}.yaml
+      run.yaml → ../.data/workflow-runs/{id}/{run-id}.yaml
 ```
 
 This structure is maintained automatically. Use `swamp repo index` to rebuild if

.gitignore

@@ -2,13 +2,13 @@ dev-logs/*
 swamp
 
 # Generated/runtime data (ignored)
-data/data/
-data/outputs/
-data/workflow-runs/
-data/inputs-evaluated/
-data/workflows-evaluated/
-data/logs/
-data/files/
+.data/data/
+.data/outputs/
+.data/workflow-runs/
+.data/inputs-evaluated/
+.data/workflows-evaluated/
+.data/logs/
+.data/files/
 
 test-repo/
 experiments/webapp/frontend/node_modules/

design/agent.md

@@ -22,7 +22,7 @@ are the recommended way to explore and understand the repository structure.
 
 ### Data Directory (Direct Access)
 
-The `/data/` directory contains the internal storage format. Agents can access
+The `/.data/` directory contains the internal storage format. Agents can access
 it directly when needed, but the layout reflects swamp's internal architecture
 rather than user-facing concerns.
 

design/expressions.md

@@ -99,11 +99,11 @@ registering custom types, functions, etc in their swamp repo.
 
 When loading the YAML, first parse the CEL expressions. Then take the data
 structures they emit and embed them in the data structure. Write those to a
-directory in the repository called `/data/inputs-evaluated/` whose structure is
-the same as `/data/inputs/`. This directory should be in a swamp repo's
+directory in the repository called `/.data/inputs-evaluated/` whose structure is
+the same as `/.data/inputs/`. This directory should be in a swamp repo's
 .gitignore file.
 
-The same is true for `/data/workflows-evaluated/`, and it should also be in
+The same is true for `/.data/workflows-evaluated/`, and it should also be in
 .gitignore.
 
 These evaluated directories are internal working directories in the data layer,

design/models.md

@@ -40,10 +40,10 @@ A model can migrate its inputs and resources from one version to the next.
 
 ## Inputs
 
-Inputs are specified as YAML files that live in the `/data/inputs/` directory of
-a repository, underneath the normalized type as a directory. The file name is
+Inputs are specified as YAML files that live in the `/.data/inputs/` directory
+of a repository, underneath the normalized type as a directory. The file name is
 `${id}.yaml`. For example,
-`data/inputs/aws/ec2/vpc/fc7fd41e-ae16-4b31-b57a-86de716e3ece.yaml`.
+`.data/inputs/aws/ec2/vpc/fc7fd41e-ae16-4b31-b57a-86de716e3ece.yaml`.
 
 The valid shape of an input is specified with a Zod 4 schema.
 
@@ -79,7 +79,7 @@ method.
 ### Logs
 
 A method may produce 0..\* log artifacts. They have a name, can have lines
-streamed to them, and by default are stored in `/data/logs/` in the repository
+streamed to them, and by default are stored in `/.data/logs/` in the repository
 underneath the normalized model type as a directory. Logs are named like
 `{model-id}-{method name}-{log name}-{timestamp}.log`.
 
@@ -91,27 +91,27 @@ artifact named 'k8slog'.
 The stream of log output should have an event emitter attached to it, so we can
 stream logs in real time.
 
-By default, the `/data/logs/` directory is not stored in git.
+By default, the `/.data/logs/` directory is not stored in git.
 
 ### Files
 
 A method may store 0..\* file artifacts. They have a name, and can be written to
-directly. By default they will be stored in the `/data/files/` directory of the
+directly. By default they will be stored in the `/.data/files/` directory of the
 repository underneath the normalized model type as a directory plus the model ID
 and method name. For example
-`data/files/aws/s3/bucket/{model-id}/{method-name}/{filename}`.
+`.data/files/aws/s3/bucket/{model-id}/{method-name}/{filename}`.
 
-By default, the `/data/files/` directory is not stored in git.
+By default, the `/.data/files/` directory is not stored in git.
 
 ## Resource
 
 Resource artifacts are used to track data about an external resource that should
 be persisted over time (for example, the data about an AWS cloud resource).
 
 A method may produce 0..1 resource as specified as YAML files that live in the
-`/data/resources/` directory of a repository, underneath the normalized type as
+`/.data/resources/` directory of a repository, underneath the normalized type as
 a directory. The file name is `${id}.yaml`. For example,
-`data/resources/aws/ec2/vpc/fc7fd41e-ae16-4b31-b57a-86de716e3ece.yaml`.
+`.data/resources/aws/ec2/vpc/fc7fd41e-ae16-4b31-b57a-86de716e3ece.yaml`.
 
 The valid shape of a resource is specified with a Zod 4 schema.
 
@@ -122,7 +122,7 @@ Resources are tracked in git.
 Data artifacts are pure data objects that are not persisted over time in git.
 
 A method may produce 0..1 data artifacts, stored as YAML files in the
-`/data/data/` directory of a repository, underneath the normalized type as a
+`/.data/data/` directory of a repository, underneath the normalized type as a
 directory. The file name is `${id}.yaml`.
 
 The valid shape of data is specified with a Zod 4 schema.
@@ -151,11 +151,11 @@ Use **data artifacts** when:
 ## Output
 
 Each method invocation produces an output record, which gets tracked in the
-`/data/outputs/` directory of a repository (which should not be tracked in git).
-The output record should track the state of the method execution, and the list
-of artifacts produced by the method. It should track state as the method
+`/.data/outputs/` directory of a repository (which should not be tracked in
+git). The output record should track the state of the method execution, and the
+list of artifacts produced by the method. It should track state as the method
 executes. It should be structured as
-`/data/outputs/{normalized-type}/{method}/{model-id}-{timestamp}.yaml`.
+`/.data/outputs/{normalized-type}/{method}/{model-id}-{timestamp}.yaml`.
 
 ## Logical Views
 
@@ -166,13 +166,13 @@ provides human/agent-friendly exploration of models by name.
 
 ```
 /models/{model-name}/
-  input.yaml      → symlink to /data/inputs/{type}/{id}.yaml
-  resource.yaml   → symlink to /data/resources/{type}/{id}.yaml
-  data.yaml       → symlink to /data/data/{type}/{id}.yaml
-  logs/           → symlink to /data/logs/{type}/{id}/
-  files/          → symlink to /data/files/{type}/{id}/
+  input.yaml      → symlink to /.data/inputs/{type}/{id}.yaml
+  resource.yaml   → symlink to /.data/resources/{type}/{id}.yaml
+  data.yaml       → symlink to /.data/data/{type}/{id}.yaml
+  logs/           → symlink to /.data/logs/{type}/{id}/
+  files/          → symlink to /.data/files/{type}/{id}/
   outputs/
-    {method}/     → symlinks to /data/outputs/{type}/{method}/{id}-*.yaml
+    {method}/     → symlinks to /.data/outputs/{type}/{method}/{id}-*.yaml
 ```
 
 This structure allows exploring all artifacts for a model in one place, using

design/repo.md

@@ -115,24 +115,24 @@ The RepoIndexService maintains two primary logical views:
 
 ```
 /models/{model-name}/
-  input.yaml → /data/inputs/{type}/{id}.yaml
-  resource.yaml → /data/resources/{type}/{id}.yaml
-  data.yaml → /data/data/{type}/{id}.yaml
-  logs/ → /data/logs/{type}/{id}/
-  files/ → /data/files/{type}/{id}/
+  input.yaml → /.data/inputs/{type}/{id}.yaml
+  resource.yaml → /.data/resources/{type}/{id}.yaml
+  data.yaml → /.data/data/{type}/{id}.yaml
+  logs/ → /.data/logs/{type}/{id}/
+  files/ → /.data/files/{type}/{id}/
   outputs/
-    {method}/ → /data/outputs/{type}/{method}/{id}-{timestamp}.yaml
+    {method}/ → /.data/outputs/{type}/{method}/{id}-{timestamp}.yaml
 ```
 
 **Workflow View (`/workflows/`):**
 
 ```
 /workflows/{workflow-name}/
-  workflow.yaml → /data/workflows/{id}.yaml
+  workflow.yaml → /.data/workflows/{id}.yaml
   runs/
     latest/ -> (points to latest timestamp)
     {timestamp}/
-      run.yaml → /data/workflow-runs/{workflow-id}/{run-id}.yaml
+      run.yaml → /.data/workflow-runs/{workflow-id}/{run-id}.yaml
       steps/
         {step-name}/
           output.yaml → symlink to step output

design/workflow.md

@@ -18,14 +18,14 @@ weighted topological sort, so thtat htye have maximum paralleism through the
 workflow. Like steps, jobs also have conditions that trigger them.
 
 Workflows are specified in YAML files, that are validated with Zod, in the
-`/data/workflows/` directory of the repository, with their `{uuid}.yaml`.
-Workflow run output is stored in `/data/workflow-runs/` at
-`/data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml`.
+`/.data/workflows/` directory of the repository, with their `{uuid}.yaml`.
+Workflow run output is stored in `/.data/workflow-runs/` at
+`/.data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml`.
 
 ## Workflow Definition
 
-Workflows are specified in `/data/workflows/{uuid}.yaml`. They have a unique id,
-a globally unique name, and a set of jobs.
+Workflows are specified in `/.data/workflows/{uuid}.yaml`. They have a unique
+id, a globally unique name, and a set of jobs.
 
 ## Jobs
 
@@ -49,7 +49,7 @@ not vary between identical inputs. (If the inputs are identical, the run order
 should be deterministic.)
 
 The output of the run will be written to a workflow run log, kept in
-`/data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml`.
+`/.data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml`.
 
 ## Logical Views
 
@@ -60,10 +60,10 @@ that provides human/agent-friendly exploration of workflows by name.
 
 ```
 /workflows/{workflow-name}/
-  workflow.yaml   → symlink to /data/workflows/{uuid}.yaml
+  workflow.yaml   → symlink to /.data/workflows/{uuid}.yaml
   runs/
     {run-id}/
-      run.yaml    → symlink to /data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml
+      run.yaml    → symlink to /.data/workflow-runs/{workflow-uuid}/{run-uuid}.yaml
       steps/
         {step-name}/
           output.yaml → symlink to step output

integration/echo_model_test.ts

@@ -122,8 +122,8 @@ Deno.test("Echo model: directory structure is correct", async () => {
     await dataRepo.save(modelType, result.data!);
 
     // Verify directory structure
-    const inputDir = join(repoDir, "data", "inputs", "swamp/echo");
-    const dataDir = join(repoDir, "data", "data", "swamp/echo");
+    const inputDir = join(repoDir, ".data", "inputs", "swamp/echo");
+    const dataDir = join(repoDir, ".data", "data", "swamp/echo");
 
     assertEquals(existsSync(inputDir), true, "Input directory should exist");
     assertEquals(

integration/repo_index_test.ts

@@ -31,14 +31,14 @@ async function withTempDir(fn: (dir: string) => Promise<void>): Promise<void> {
 
 async function setupRepoDir(dir: string): Promise<void> {
   const subdirs = [
-    "data/inputs",
-    "data/resources",
-    "data/data",
-    "data/outputs",
-    "data/workflows",
-    "data/workflow-runs",
-    "data/logs",
-    "data/files",
+    ".data/inputs",
+    ".data/resources",
+    ".data/data",
+    ".data/outputs",
+    ".data/workflows",
+    ".data/workflow-runs",
+    ".data/logs",
+    ".data/files",
     "models",
     "workflows",
   ];
@@ -318,7 +318,7 @@ Deno.test("Integration: repo index verify detects broken symlinks", async () =>
     const modelDir = join(repoDir, "models", "broken-model");
     await ensureDir(modelDir);
     await Deno.symlink(
-      "../data/inputs/nonexistent/file.yaml",
+      "../.data/inputs/nonexistent/file.yaml",
       join(modelDir, "input.yaml"),
     );
 
@@ -341,7 +341,7 @@ Deno.test("Integration: repo index prune removes broken symlinks", async () => {
     await ensureDir(modelDir);
     const brokenLink = join(modelDir, "input.yaml");
     await Deno.symlink(
-      "../data/inputs/nonexistent/file.yaml",
+      "../.data/inputs/nonexistent/file.yaml",
       brokenLink,
     );
 
@@ -453,7 +453,7 @@ Deno.test("CLI: repo index --verify fails on broken symlinks", async () => {
     const modelDir = join(repoDir, "models", "broken-verify-test");
     await ensureDir(modelDir);
     await Deno.symlink(
-      "../data/inputs/nonexistent/file.yaml",
+      "../.data/inputs/nonexistent/file.yaml",
       join(modelDir, "input.yaml"),
     );
 
@@ -481,7 +481,7 @@ Deno.test("CLI: repo index --prune removes broken symlinks", async () => {
     await ensureDir(modelDir);
     const brokenLink = join(modelDir, "input.yaml");
     await Deno.symlink(
-      "../data/inputs/nonexistent/file.yaml",
+      "../.data/inputs/nonexistent/file.yaml",
       brokenLink,
     );
 

integration/workflow_test.ts

@@ -1121,7 +1121,7 @@ Deno.test("CLI: model delete cleans up empty type directories", async () => {
     assertEquals(createResult.code, 0, "Model create should succeed");
 
     // Verify the directory structure was created
-    const inputsDir = `${repoDir}/data/inputs`;
+    const inputsDir = `${repoDir}/.data/inputs`;
     const echoDir = `${inputsDir}/swamp/echo`;
     const swampDir = `${inputsDir}/swamp`;
 

src/domain/repo/repo_index_service.ts

@@ -50,22 +50,22 @@ export interface RebuildResult {
  *
  * Model View (`/models/{model-name}/`):
  * ```
- * input.yaml      → ../data/inputs/{type}/{id}.yaml
- * resource.yaml   → ../data/resources/{type}/{id}.yaml
- * data.yaml       → ../data/data/{type}/{id}.yaml
- * logs/           → ../data/logs/{type}/{id}/
- * files/          → ../data/files/{type}/{id}/
+ * input.yaml      → ../.data/inputs/{type}/{id}.yaml
+ * resource.yaml   → ../.data/resources/{type}/{id}.yaml
+ * data.yaml       → ../.data/data/{type}/{id}.yaml
+ * logs/           → ../.data/logs/{type}/{id}/
+ * files/          → ../.data/files/{type}/{id}/
  * outputs/
- *   {method}/     → ../data/outputs/{type}/{method}/
+ *   {method}/     → ../.data/outputs/{type}/{method}/
  * ```
  *
  * Workflow View (`/workflows/{workflow-name}/`):
  * ```
- * workflow.yaml   → ../data/workflows/workflow-{id}.yaml
+ * workflow.yaml   → ../.data/workflows/workflow-{id}.yaml
  * runs/
  *   latest/       → {latest-timestamp}/
  *   {timestamp}/
- *     run.yaml    → ../data/workflow-runs/{workflow-id}/workflow-run-{run-id}.yaml
+ *     run.yaml    → ../.data/workflow-runs/{workflow-id}/workflow-run-{run-id}.yaml
  *     steps/
  *       {step-name}/
  *         output.yaml → symlink to step output
@@ -161,7 +161,7 @@ export interface RepoIndexService {
    * Rebuilds all logical views from scratch.
    *
    * Deletes existing /models/ and /workflows/ directories
-   * and recreates them from the /data/ directory.
+   * and recreates them from the /.data/ directory.
    *
    * @returns Rebuild result with counts
    */

src/domain/repo/repo_service.ts

@@ -205,7 +205,7 @@ Use \`swamp --help\` to see available commands.
   private async createDataDirectoryStructure(
     repoPath: RepoPath,
   ): Promise<void> {
-    const dataDir = join(repoPath.value, "data");
+    const dataDir = join(repoPath.value, ".data");
     const subdirs = [
       "inputs",
       "resources",