feat(files): folder operations for the File block, and folders as scope by mzxchandra · Pull Request #7377 · simstudioai/sim

mzxchandra · 2026-09-02T00:44:19Z

What

Adds folder operations to file_v5 and to the agent tool surface, and makes a folder a scope on the file operations that already existed rather than a second set of operations beside them.

New operations: List, Create Folder, Move Folder, Delete Folder, Restore Folder, Move File.

List answers "what is in here" — subfolders and files together, direct children by default, the whole subtree under Recursive, subject to Max Depth and Search. Entries are a discriminated union on kind, so a consumer narrows before reaching for the fields only one side has. The listing is capped with a truncated flag rather than unbounded, now that it includes files.

Read / Get Content / Compress / Append gain an optional Folder above their file picker. It narrows what the picker offers; on the three read operations it also stands for that folder's files when none are picked, resolved when the workflow runs so a file added later is included. Append only narrows — a folder is not something you can append to, so it shapes the options and does not travel.

Write gains a folder destination, above File Name, because it names where before it names what.

Where to spend review attention

Path handling — this is where the bugs live. Two spellings circulate: the stored display path, which backslash-escapes a slash inside a folder name (Reports/Q3\/Q4), and the canonical percent-encoded path the tools take (/Reports/Q3%2FQ4). A folder genuinely named Q3/Q4 is one level in both and two if either is split on /. folderPathSegments picks the parser by the leading slash the canonical form always carries; resolveFolderIdsForPaths, isFileInFolderScope and selectDirectoryEntries are pure and tested against exactly that case. I shipped this bug once during development — it surfaced as "Workspace file folder path contains an empty name" on a correctly picked folder.

Delete Folder's recursive flag is a guard, not a scope. Without it, deleting a non-empty folder fails, and it is user-only so a model asked to "clean up" a folder cannot set it on a guess. Deliberately unlike the read family's Include Subfolders, which is a scope and defaults on.

Canonical pairs. Every folder field is a single-select tree paired with a manual entry. The pair is load-bearing, not a convenience: < reference autocomplete comes from TagDropdown, which scans for the last < before the cursor, so it can only live on a text surface. A checkbox tree has no cursor.

No new tools for folder-scoped reads. The manage operation already accepts folderPaths alongside file ids, so the folder scope reuses file_read / file_get_content / file_compress. Two nearly-overlapping tools are harder for an agent to choose between than one that takes either.

Root is the absence of a selection, not a row — a root row would be meaningless on create and wrong on delete. The placeholders say so.

Known characteristic, not a blocker

The list case reads the workspace's full folder and file lists and filters in memory, the same way expandFolderPathsToFileIds and roughly ten existing callers of listAllWorkspaceFiles do. Output is capped by limit with a truncated flag, but the read itself is not bounded. queryWorkspaceFilePage is the paged, folder-scoped alternative if this becomes a problem on large workspaces — worth doing as a follow-up across all callers rather than only this one.

Testing

Scoped suites green: 1918 tests across blocks, tools/file, lib/internal/file, lib/workspace-files, lib/workflows/subblocks.
All 44 repo audits pass, including check-canvas-sentences, generated docs, and tool metadata.
check:api-validation and check:client-boundary clean; biome clean.
Verified in the browser on a local workspace: folder picking, create folder, write into a folder, and the canvas card label.
Pure path/selection logic (folder-path-selection, directory-listing, display) has direct unit coverage including the slash-in-name case.

Follow-ups (not in this PR)

Knowledge and Tables folder tools — same shapes, stacked on this. Their application layers and v2 routes already exist in staging; what they need is agent-callable tools, executor admission on their folder operations, a dispatch path, and block wiring.
Knowledge and table folder trees have no realtime broadcast, so they wait out their stale time; only workspace files invalidate live.

vercel · 2026-09-02T00:44:24Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
docs	Ready	Preview	Sep 2, 2026 5:39am UTC

greptile-apps · 2026-09-02T00:46:56Z

Greptile Summary

The PR adds folder creation, movement, deletion, restoration, and listing to File v5 while extending existing file operations with folder scopes.

Adds canonical folder-path parsing, selection, and directory-listing utilities.
Adds folder-aware File block inputs, tools, contracts, execution paths, and generated metadata.
Adds a folder tree selector and scopes workspace-file pickers to the selected folder.
Adds tests and documentation for folder operations and encoded path behavior.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
apps/sim/blocks/blocks/file.ts	Extends File v5 configuration and parameter transformation with folder operations and folder-scoped file behavior.
apps/sim/lib/internal/file/operations.ts	Implements internal dispatch for the new folder and file-move operations.
apps/sim/lib/workspace-files/application/workspace-file-folders.ts	Adds application-layer folder lifecycle operations for workspace files.
apps/sim/lib/workspace-files/directory-listing.ts	Adds bounded folder-and-file directory listing with recursive, depth, and search controls.
apps/sim/app/workspace/[workspaceId]/w/[workflowId]/components/panel/components/editor/components/sub-block/components/sim-folder-tree-selector/sim-folder-tree-selector.tsx	Adds the single-select folder tree used by File block folder parameters.
apps/sim/lib/api/contracts/tools/file.ts	Extends File tool contracts with folder operations and scoped file inputs.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  UI[File v5 block inputs] --> Params[Parameter transformation]
  Params --> Tool[File tool dispatch]
  Tool --> Ops[Authorized file operations]
  Ops --> Folders[Workspace folders]
  Ops --> Files[Workspace files]
  Folders --> Result[Folder or directory result]
  Files --> Result

_{Reviews (5): Last reviewed commit: "fix(files): resolve an overwrite target ..." | Re-trigger Greptile}

cubic-dev-ai

All reported issues were addressed across 42 files

_{Heads up: you’re close to your included review allowance. Set a flex budget so reviews don’t pause.

Re-trigger cubic}

Adds folder operations to file_v5 and to the agent tool surface, and makes a folder a scope on the file operations that already existed rather than a second set of operations beside them. New operations: List, Create Folder, Move Folder, Delete Folder, Restore Folder, and Move File. List answers "what is in here" — subfolders and files together, direct children by default, the whole subtree under Recursive, subject to Max Depth and Search. Its entries are a discriminated union on kind, so a consumer narrows before reaching for the fields only one side has, and the listing is capped with a truncated flag rather than being unbounded now that it includes files. Read, Get Content, Compress and Append gain an optional Folder above their file picker. It narrows what the picker offers, and on the three read operations it also stands for that folder's files when none are picked — resolved when the workflow runs, so a file added later is included. Append only narrows: a folder is not something you can append to, so it shapes the options and does not travel. Write gains a folder destination, placed above File Name because it names where before it names what. Every folder field is one single-select tree control paired with a manual entry. The picker cannot hold a reference expression — `<` autocomplete comes from TagDropdown, which scans for the last `<` before the cursor and so only exists on a text surface — which is why the pair exists rather than being a convenience. Root is the absence of a selection rather than a row, since a root row would be meaningless on create and wrong on delete. Path handling is the part most worth reviewing. Two spellings circulate: the stored display path, which backslash-escapes a slash inside a folder name, and the canonical percent-encoded path the tools take. A folder genuinely named "Q3/Q4" is one level in both and two if either is split on "/", so folderPathSegments decodes by the leading slash the canonical form carries, and resolveFolderIdsForPaths, isFileInFolderScope and selectDirectoryEntries are pure and tested against exactly that case. Delete Folder keeps its recursive flag as a guard rather than a scope: without it, deleting a non-empty folder fails, and it is user-only so a model asked to clean up a folder cannot set it on a guess.

…tools

The list operation's entry selection had no tests. These pin the parts that are easy to get subtly wrong: a file counts one level below its folder, so the deepest folder a depth admits arrives without its contents; search filters the result rather than the traversal, so a deep match still reports even when its parent folders do not match; and a cut listing reports truncated rather than looking complete.

Three fixes from the AI reviewers on #7377. The single-file picker kept its unscoped options when the folder changed: singleFileOptions read scopedWorkspaceFiles but listed workspaceFiles in its memo dependencies, so it never recomputed. That is the picker Append uses, so the folder narrowing I added for it did nothing after the first render. The v2 files-folders route called toWorkspaceFileFolderPathView while importing staging's toV2Folder — a conflict resolution I got half right, and a compile error I would have found by running the type-check I skipped. The file is back to staging's version, since keeping its projection was the decision anyway. fileFamilyInput took Record<string, any>, which the repo's TypeScript rules forbid. It now takes an interface naming the three fields it actually reads, so a shape change fails at the call site rather than at run time.

cubic-dev-ai

All reported issues were addressed

_{Heads up: you’re close to your included review allowance. Set a flex budget so reviews don’t pause.

Re-trigger cubic}

Three P1s from cubic, all real, all mine, and all introduced by the rebase. Folder expansion was gone from read, get content and compress. Taking the upstream copy of the operations file and re-applying only the folder CASES dropped expandFolderPathsToFileIds and the folderPaths handling with it, so the block sent a folder, the contract accepted it, the tool forwarded it, and the operation ignored it — a folder-only selection failed with "File is required". file_write never read folderPath: it derived the destination from the file name alone, so choosing a folder wrote to the workspace root. It now prefixes the picked folder's decoded segments onto whatever the name itself spells. The six folder tool ids were registered in FILE_MANAGE_TOOL_IDS, which chooses the handler branch, but not in FILE_TOOL_IDS, which decides whether the internal registry has a handler at all. Every workflow execution of them would have failed before reaching the branch. None of this was caught because the tests covered the two ends — the block's params and the pure path helpers — and nothing covered the middle. Worse, the test file's module mock stubbed one export from workspace-file-folders, so the five folder operations were undefined and no test could reach a folder case at all. The mock is complete now, and five tests cover the wiring: a folder-only read expands, a missing folder refuses rather than reading nothing, write lands in the chosen folder and still lands at the root without one, and list returns folders and files together.

…lock Two more from cubic, both correct. The workflow diff rendered every folder change as the same generic label. Adding sim-folder-tree-selector to SELECTOR_TYPES_HYDRATION_REQUIRED without a resolver makes each consumer of that list fall back to a semantic placeholder — the canvas card showed "-" for the same reason, which I fixed there and not here. The comparison path now calls the same resolveFolderPathLabel, so a diff between two folders shows which two. Max Depth and Limit were bounded only at the contract, so an out-of-range value was accepted while configuring and failed on the first run. They now carry the same ceilings in the block, where the mistake is made.

mzxchandra · 2026-09-02T02:02:38Z

@greptile review

mzxchandra · 2026-09-02T02:02:43Z

@cubic review

cubic-dev-ai · 2026-09-02T02:02:49Z

@cubic review

@mzxchandra I can't start this review because your workspace has reached its monthly review limit. cubic has reviewed 574,058 of the 560,000 allowed lines of code this month. Reviews resume on 28 September 2026 (in 27 days). Enable flex capacity to cover overages automatically and resume reviews now. Learn how flex capacity works.

To help optimise your usage, you can tune cubic to get the most out of your usage limits:

Stop reviewing test files
Turn off incremental reviews
Use the cubic CLI locally
Skip non-default base branches

Learn more →

…tools # Conflicts: # apps/sim/tools/generated/tool-metadata.ts

mzxchandra · 2026-09-02T02:09:06Z

@cubic review

cubic-dev-ai · 2026-09-02T02:09:11Z

@cubic review

@mzxchandra I have started the AI code review. It will take a few minutes to complete.

cubic-dev-ai

All reported issues were addressed across 43 files

_{Fix all with cubic | Re-trigger cubic}

CI's Lint and Test failed on sub-block-password.test.tsx, a test this branch does not touch, with a missing icon export from a partial mock. The cause was a deep import. sub-block.tsx takes every other subblock component from the components barrel, and that test mocks the barrel wholesale — which is what keeps the editor's module graph out of its way. My folder picker was imported directly from its file instead, bypassing the mock and pulling the real chain in behind it: use-resource-folders, the realtime invalidation room, the socket provider, the workflow stores, and through them the block and trigger registries. Every trigger's icon then had to appear in that test's icon mock. Exporting the picker from the barrel and importing it from there fixes it with no change to the test, which is the sign it was the right fix — I had first patched the mock, then stubbed the socket provider, and both were treating a symptom. The repo's import rules already say to use the barrel; I had not.

mzxchandra · 2026-09-02T02:20:00Z

@cubic review

cubic-dev-ai · 2026-09-02T02:20:05Z

@cubic review

@mzxchandra I have started the AI code review. It will take a few minutes to complete.

mzxchandra · 2026-09-02T02:20:12Z

@greptile review

cubic-dev-ai

No issues found across 44 files

Confidence score: 5/5

Automated review surfaced no issues in the provided summaries.
No files require special attention.

_{Re-trigger cubic}

cubic raised this twice; I deferred it once and it deserved the fix. Append resolves a picked file by canonical id, where the folder beside it is redundant. The advanced entry supplies a name instead, and a name is only unique within a folder — a workspace-wide lookup takes the oldest match anywhere, so the folder sat next to the field looking like it scoped the operation while doing nothing. That is the same objection that had me remove this field earlier in the branch, reappearing in advanced mode only. The rule is now stated properly: the folder travels when it is what identifies the file, and stays behind when the id already does. A named append expands the chosen folder through the authorized use case, matches the name inside it, and resolves by the id it finds — so no path-shaped reference is ever built and the slash-in-a-folder-name hazard cannot arise on this path. A name that is not in the folder refuses rather than reaching for a same-named file elsewhere. Tests cover the case that motivated it: two files named notes.md in different folders, appending to the one in the folder that was picked.

…tools # Conflicts: # apps/sim/lib/workflows/subblocks/display.ts

A slash-delimited reference cannot express a folder whose own name contains a slash. `Q3/Q4` joins to two segments and re-reads as two levels, so the overwrite lookup never found the existing file and the write landed as a duplicate beside it. The guard on folder id and name meant it could not overwrite the WRONG file, which is why this was bounded rather than dangerous. I introduced it: before this branch the segments came only from splitting the file name, so no segment could contain a slash. Prepending the picked folder's decoded segments is what made one able to. When a folder was picked its canonical path is unambiguous, so the target is found inside that folder by name and resolved by id — the same route the named append now takes, using the same authorized expansion. A typed path with no picked folder keeps the joined reference, which is exactly as expressive as the name it came from.

mzxchandra · 2026-09-02T05:39:25Z

@greptile

mzxchandra · 2026-09-02T05:41:11Z

@cubic review

cubic-dev-ai · 2026-09-02T05:41:16Z

@cubic review

@mzxchandra I have started the AI code review. It will take a few minutes to complete.

cubic-dev-ai

6 issues found across 44 files

Confidence score: 2/5

apps/sim/lib/internal/file/operations.ts: scoped.find can select an arbitrary duplicate filename in a recursive scope, causing edits to the wrong file; reject ambiguous names or require an exact path/ID.
apps/sim/lib/internal/file/operations.ts: manually entered filenames beginning with wf_ can be misclassified as file IDs, bypassing folder lookup and targeting the wrong resource; distinguish canonical IDs from user-provided names.
apps/sim/lib/api/contracts/tools/file.ts: nested file_create_folder paths are accepted even though execution fails when the parent is absent, while path-based delete_folder audit entries omit resourceId; align parent creation with execution and preserve the resolved folder ID in auditing.
apps/sim/lib/workflows/subblocks/display.ts: folder selections can still render as -, and legacy array-shaped values remain unnamed in previews; apply the same folder-path normalization and resolveFolderPathLabel handling across editor and preview resolution.

Prompt for AI agents (unresolved issues)


Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="apps/sim/lib/internal/file/operations.ts">

<violation number="1" location="apps/sim/lib/internal/file/operations.ts:1235">
P2: When a manually entered filename starts with `wf_`, this condition treats it as a file id and skips the folder lookup. Distinguish canonical picked ids from names without reserving a valid filename prefix, so named appends remain scoped.</violation>

<violation number="2" location="apps/sim/lib/internal/file/operations.ts:1242">
P1: When a recursive folder scope contains duplicate filenames, `scoped.find` appends to whichever file happens to come first. Reject ambiguous names or require an exact path/id instead of silently modifying an arbitrary descendant.</violation>
</file>

<file name="apps/sim/lib/workflows/subblocks/display.ts">

<violation number="1" location="apps/sim/lib/workflows/subblocks/display.ts:619">
P2: The read-only preview still shows `-` for selected `sim-folder-tree-selector` values. Wire `resolveFolderPathLabel` into the preview resolver's hydrated-name chain so the editor and preview display folder scopes consistently.

(Based on your team's feedback about resolving hydrated folder selectors in every display consumer.)</violation>

<violation number="2" location="apps/sim/lib/workflows/subblocks/display.ts:624">
P3: When a workflow contains a legacy array-shaped folder value, this resolver cannot name it even though the selector and file block still normalize it. Apply the same folder-path normalization before parsing so migrated folder selections remain readable.</violation>
</file>

<file name="apps/sim/lib/api/contracts/tools/file.ts">

<violation number="1" location="apps/sim/lib/api/contracts/tools/file.ts:243">
P2: When the new path-based `delete_folder` operation succeeds, its `FOLDER_DELETED` audit entry has no `resourceId` because the projector reads `input.folderId`, which this contract never supplies. Resolve and retain the folder ID for auditing while continuing to return the public path.

(Based on your team's feedback about preserving canonical folder IDs in folder deletion audits.)</violation>

<violation number="2" location="apps/sim/lib/api/contracts/tools/file.ts:243">
P2: When `file_create_folder` receives a nested path whose parent does not exist, this contract accepts it but execution returns `Parent folder not found`, contradicting the tool's parent-creation behavior. Route this operation through the folder-path materializer so missing parents are created before the leaf.</violation>
</file>

_{Re-trigger cubic}

cubic-dev-ai · 2026-09-02T05:51:20Z

+            folderPaths: [folderPath],
+            includeSubfolders,
+          })
+          const match = scoped.find((file) => file.name === fileName)


P1: When a recursive folder scope contains duplicate filenames, scoped.find appends to whichever file happens to come first. Reject ambiguous names or require an exact path/id instead of silently modifying an arbitrary descendant.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/internal/file/operations.ts, line 1242: <comment>When a recursive folder scope contains duplicate filenames, `scoped.find` appends to whichever file happens to come first. Reject ambiguous names or require an exact path/id instead of silently modifying an arbitrary descendant.</comment> <file context> @@ -1090,14 +1220,37 @@ export async function executeFileManageOperation( + folderPaths: [folderPath], + includeSubfolders, + }) + const match = scoped.find((file) => file.name === fileName) + if (!match) { + throw new OrchestrationError('not_found', `No file named ${fileName} in ${folderPath}`) </file context>

cubic-dev-ai · 2026-09-02T05:51:20Z

+         * hazard of a path-shaped reference never arises.
+         */
+        let scopedReference = fileName
+        if (folderPath && !fileName.startsWith('wf_')) {


P2: When a manually entered filename starts with wf_, this condition treats it as a file id and skips the folder lookup. Distinguish canonical picked ids from names without reserving a valid filename prefix, so named appends remain scoped.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/internal/file/operations.ts, line 1235: <comment>When a manually entered filename starts with `wf_`, this condition treats it as a file id and skips the folder lookup. Distinguish canonical picked ids from names without reserving a valid filename prefix, so named appends remain scoped.</comment> <file context> @@ -1090,14 +1220,37 @@ export async function executeFileManageOperation( + * hazard of a path-shaped reference never arises. + */ + let scopedReference = fileName + if (folderPath && !fileName.startsWith('wf_')) { + const scoped = await expandFolderPathsToFiles({ + principal, </file context>

cubic-dev-ai · 2026-09-02T05:51:20Z

+ * request per canvas row, no loading state, and nothing to go stale that the
+ * stored path has not gone stale with.
+ */
+export function resolveFolderPathLabel(


P2: The read-only preview still shows - for selected sim-folder-tree-selector values. Wire resolveFolderPathLabel into the preview resolver's hydrated-name chain so the editor and preview display folder scopes consistently.

(Based on your team's feedback about resolving hydrated folder selectors in every display consumer.)

View Feedback

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/workflows/subblocks/display.ts, line 619: <comment>The read-only preview still shows `-` for selected `sim-folder-tree-selector` values. Wire `resolveFolderPathLabel` into the preview resolver's hydrated-name chain so the editor and preview display folder scopes consistently. (Based on your team's feedback about resolving hydrated folder selectors in every display consumer.) </comment> <file context> @@ -602,3 +603,31 @@ export function resolveSandboxLabel( + * request per canvas row, no loading state, and nothing to go stale that the + * stored path has not gone stale with. + */ +export function resolveFolderPathLabel( + subBlock: SubBlockConfig | undefined, + rawValue: unknown </file context>

cubic-dev-ai · 2026-09-02T05:51:20Z

+export const fileManageCreateFolderBodySchema = z.object({
+  operation: z.literal('create_folder'),
+  workspaceId: z.string().min(1).optional(),
+  path: v2NonRootFolderPathInputSchema,


P2: When the new path-based delete_folder operation succeeds, its FOLDER_DELETED audit entry has no resourceId because the projector reads input.folderId, which this contract never supplies. Resolve and retain the folder ID for auditing while continuing to return the public path.

(Based on your team's feedback about preserving canonical folder IDs in folder deletion audits.)

View Feedback

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/api/contracts/tools/file.ts, line 243: <comment>When the new path-based `delete_folder` operation succeeds, its `FOLDER_DELETED` audit entry has no `resourceId` because the projector reads `input.folderId`, which this contract never supplies. Resolve and retain the folder ID for auditing while continuing to return the public path. (Based on your team's feedback about preserving canonical folder IDs in folder deletion audits.) </comment> <file context> @@ -137,6 +208,61 @@ export const fileManageDecompressBodySchema = z +export const fileManageCreateFolderBodySchema = z.object({ + operation: z.literal('create_folder'), + workspaceId: z.string().min(1).optional(), + path: v2NonRootFolderPathInputSchema, +}) + </file context>

cubic-dev-ai · 2026-09-02T05:51:20Z

+export const fileManageCreateFolderBodySchema = z.object({
+  operation: z.literal('create_folder'),
+  workspaceId: z.string().min(1).optional(),
+  path: v2NonRootFolderPathInputSchema,


P2: When file_create_folder receives a nested path whose parent does not exist, this contract accepts it but execution returns Parent folder not found, contradicting the tool's parent-creation behavior. Route this operation through the folder-path materializer so missing parents are created before the leaf.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/api/contracts/tools/file.ts, line 243: <comment>When `file_create_folder` receives a nested path whose parent does not exist, this contract accepts it but execution returns `Parent folder not found`, contradicting the tool's parent-creation behavior. Route this operation through the folder-path materializer so missing parents are created before the leaf.</comment> <file context> @@ -137,6 +208,61 @@ export const fileManageDecompressBodySchema = z +export const fileManageCreateFolderBodySchema = z.object({ + operation: z.literal('create_folder'), + workspaceId: z.string().min(1).optional(), + path: v2NonRootFolderPathInputSchema, +}) + </file context>

cubic-dev-ai · 2026-09-02T05:51:20Z

+  rawValue: unknown
+): string | null {
+  if (subBlock?.type !== 'sim-folder-tree-selector') return null
+  if (typeof rawValue !== 'string' || rawValue.trim() === '') return null


P3: When a workflow contains a legacy array-shaped folder value, this resolver cannot name it even though the selector and file block still normalize it. Apply the same folder-path normalization before parsing so migrated folder selections remain readable.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/sim/lib/workflows/subblocks/display.ts, line 624: <comment>When a workflow contains a legacy array-shaped folder value, this resolver cannot name it even though the selector and file block still normalize it. Apply the same folder-path normalization before parsing so migrated folder selections remain readable.</comment> <file context> @@ -602,3 +603,31 @@ export function resolveSandboxLabel( + rawValue: unknown +): string | null { + if (subBlock?.type !== 'sim-folder-tree-selector') return null + if (typeof rawValue !== 'string' || rawValue.trim() === '') return null + + try { </file context>

mzxchandra · 2026-09-02T06:44:03Z

Superseded by #7388 — same change, one commit against current staging.

Closing rather than continuing here because the incremental history was feeding the review loop: 11 commits and two staging merges meant each round reviewed a diff-of-diffs, and the last round's findings were mostly bugs in the fixes from the round before. #7388 is byte-identical in net diff (verified) plus six further fixes from this PR's final round.

The reasoning argued on the threads here is carried into #7388's description rather than lost — the two path spellings and why resolution goes by id, delete's recursive being a guard rather than a scope, the executor widening and its per-tool justification, and the known-cost items (the unbounded workspace read behind list, and recursive folder listing being unreachable in staging's knowledge/table public schemas).

mzxchandra marked this pull request as draft September 2, 2026 00:44

greptile-apps Bot reviewed Sep 2, 2026

View reviewed changes

Comment thread apps/sim/blocks/blocks/file.ts

vercel Bot deployed to Preview September 2, 2026 00:50 View deployment

cubic-dev-ai Bot reviewed Sep 2, 2026

View reviewed changes

mzxchandra added 3 commits September 1, 2026 18:14

Merge remote-tracking branch 'origin/staging' into feat/agent-folder-…

ef927b1

…tools

mzxchandra force-pushed the agent-folder-tools branch from e3fb656 to b5a77e3 Compare September 2, 2026 01:34

mzxchandra marked this pull request as ready for review September 2, 2026 01:35

vercel Bot deployed to Preview September 2, 2026 01:37 View deployment

vercel Bot temporarily deployed to Preview September 2, 2026 01:41 Inactive

cubic-dev-ai Bot reviewed Sep 2, 2026

View reviewed changes

vercel Bot temporarily deployed to Preview September 2, 2026 01:56 Inactive

vercel Bot temporarily deployed to Preview September 2, 2026 02:00 Inactive

Merge remote-tracking branch 'origin/staging' into feat/agent-folder-…

0878ae6

…tools # Conflicts: # apps/sim/tools/generated/tool-metadata.ts

vercel Bot deployed to Preview September 2, 2026 02:10 View deployment

cubic-dev-ai Bot reviewed Sep 2, 2026

View reviewed changes

Comment thread apps/sim/blocks/blocks/file.ts

Comment thread apps/sim/lib/internal/file/operations.ts

vercel Bot temporarily deployed to Preview September 2, 2026 02:17 Inactive

cubic-dev-ai Bot reviewed Sep 2, 2026

View reviewed changes

mzxchandra added 3 commits September 1, 2026 22:31

Merge remote-tracking branch 'origin/staging' into feat/agent-folder-…

94176fc

…tools # Conflicts: # apps/sim/lib/workflows/subblocks/display.ts

vercel Bot temporarily deployed to Preview September 2, 2026 05:39 Inactive

simstudioai deleted a comment from cursor Bot Sep 2, 2026

cubic-dev-ai Bot reviewed Sep 2, 2026

View reviewed changes

mzxchandra mentioned this pull request Sep 2, 2026

feat(files): folder operations for the File block, and folders as scope #7388

Open

mzxchandra closed this Sep 2, 2026

Uh oh!

Conversation

mzxchandra commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Where to spend review attention

Known characteristic, not a blocker

Testing

Follow-ups (not in this PR)

Uh oh!

vercel Bot commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

greptile-apps Bot commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Flowchart

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot commented Sep 2, 2026

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot commented Sep 2, 2026

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

mzxchandra commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot commented Sep 2, 2026

Uh oh!

cubic-dev-ai Bot left a comment

Choose a reason for hiding this comment

mzxchandra commented Sep 2, 2026 •

edited

Loading

vercel Bot commented Sep 2, 2026 •

edited

Loading

greptile-apps Bot commented Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot left a comment •

edited

Loading

cubic-dev-ai Bot left a comment •

edited

Loading

cubic-dev-ai Bot left a comment •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading

cubic-dev-ai Bot Sep 2, 2026 •

edited

Loading