Fix char-boundary panic in WorkflowDataSource preview truncation (APP-5287)

[warp-agent-staging]

Description

Fixes a panic in the AI @ context-menu Workflow data source: WorkflowDataSource::run_query built a hover-preview description for each workflow by slicing its content at a hard-coded byte offset (&content_preview[..197]) whenever the preview exceeded 200 bytes. If a multi-byte UTF-8 character (emoji, CJK, accented Latin, etc.) straddled byte 197, this slice panicked with byte index 197 is not a char boundary.

run_query reads workflows from CloudModel (Warp Drive), not from local workflow YAML, so the panic is triggered by workflow content synced from Warp Drive, not by the user's typed query. It fires when the Workflows category is opened in the @ context menu (i.e. AIContextMenuAction::CategorySelected { category: Workflows }) or when a query runs against it — not on the bare @ zero state before a category is selected.

Note: the initial report speculated this was a ByteOffset/CharOffset (crates/string-offset) misuse, but that crate is not on this code path — the actual bug is the raw byte slice shown above.

Linked Issue

Fixes APP-5287

Changes

• app/src/search/ai_context_menu/workflows/data_source.rs: replace the raw &content_preview[..197] slice with the existing safe_truncate helper (app/src/search/ai_context_menu/mod.rs), which floors to the nearest valid UTF-8 char boundary. This is the same helper already used by the sibling code/notebooks/rules/workflows search_item.rs renderers, so all ai_context_menu truncation now goes through one path.

I audited the other ai_context_menu data sources (blocks, rules, skills, conversations, notebooks) for the same raw-byte-slice pattern; none of them have it — notebooks/data_source.rs already used a char_indices-based safe truncation, and the search_item.rs renderers already used safe_truncate. Only workflows/data_source.rs had the bug, so the fix is scoped to that one line.

Testing

This is a one-line logic fix with no unit test attached; coverage instead comes from the GUI verification below, run on an equivalent fix branch:

• cargo nextest run -p warp -E 'test(search::ai_context_menu)' — passing, no regressions.
• cargo clippy -p warp --all-targets --tests -- -D warnings — clean.
• ./script/format — clean (no changes needed beyond the edit).

Screenshots / Videos

Captured on an equivalent fix branch (same repro workflow, same machine):

• Fixed branch: @ → Workflows → the repro workflow renders with a truncated preview, filtering works, and the app survives: https://oz.staging.warp.dev/artifacts/019ff146-9d8f-72ee-b741-b9b6f911078f
• master (cd49bd7), same machine and workflow: the window vanishes on opening Workflows (the panic): https://oz.staging.warp.dev/artifacts/019ff14c-2725-7821-91a7-b1973b08c5c0
• Client log from that crash run:

  [ERROR] [panic] thread 'main' panicked at 'byte index 197 is not a char boundary; it is inside '🚀' (bytes 195..199) of `aaaa…🚀-BOUNDARY-TAIL-PADDING-TO-EXCEED-200-BYTES`': app/src/search/ai_context_menu/workflows/data_source.rs:69
  

  fired right after AIContextMenuAction::CategorySelected { category: Workflows }.
• Stills: https://oz.staging.warp.dev/artifacts/019ff145-74c6-73ad-8c15-0a3654dcff1a and https://oz.staging.warp.dev/artifacts/019ff14c-6751-7e40-88f5-608f448b1f92

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

CHANGELOG-BUG-FIX: Fixed a crash in the AI @ context menu when a workflow's content included multi-byte characters (emoji, CJK, accented Latin) near the preview truncation boundary.

Conversation: https://staging.warp.dev/conversation/b38216ff-032f-47cf-96bf-c7b058b89806
Run: https://oz.staging.warp.dev/runs/019ff0fd-e7e7-73c8-a339-b9ef2471b442

This PR was generated with Oz.

[warp-agent-staging]

Heads up: this PR and #14935 are two independent fixes for the same panic (APP-5287), produced by two parallel agent runs. Both replace the raw &content_preview[..197] slice with safe_truncate and add a multibyte regression test. Only one should be merged.

Differences worth weighing:

• This PR is out of draft and adds a second short_content_is_not_truncated test plus a wider audit of the other ai_context_menu data sources.
• [APP-5287] Fix char-boundary panic in workflow AI context menu preview #14935 is still draft but carries visual proof from a real GUI run: a recording of the @ menu working on the fixed branch, a contrast recording of master crashing on the same click, and the client log with the verbatim panic at data_source.rs:69.
• One factual correction that applies to this PR's description: the panic does not fire on the bare @ zero state. Verification on a real build showed the crash fires when the Workflows category is opened, or when a query runs against WorkflowDataSource. Also note run_query reads workflows from CloudModel/Warp Drive, not from local workflow YAML, which matters for anyone reproducing this by hand.

[vorporeal]

hey @warp-agent can you address my comment, and close the other PR?

[warp-local-for-testing-only]

Your GitHub account is not connected to Warp. Connect it here.

[warp-local-for-testing-only]

Oz couldn't complete this task because required inputs, permissions, or resources were invalid or unavailable.

Error Details:
Failed to execute task: container exited with non-zero status: 1

[vorporeal]

hey @warp-agent can you address my review comment, and close the other PR?

[warp-local-for-testing-only]

Your GitHub account is not connected to Warp. Connect it here.