fix(codeRefs): replace exec() with execFile() to prevent command injection

[unwakuche-ld]

Here is a PR description that matches what the branch now ships. I have not applied it to the PR, since that is a remote action.

Summary

Fixes a command injection vulnerability in Code References alias generation (bug bounty report).

Vector: FlagAliases.generateCsv and FlagAliases.codeRefsVersionCheck in src/providers/codeRefs.ts built a single command string and ran it through child_process.exec, which spawns a shell. The string interpolated workspace-controlled values inside double quotes:

• intConfig.project, from launchdarkly.project in a repo's .vscode/settings.json
• directory, the workspace folder path (cloned folder name or a .code-workspace path)

A repository could therefore run arbitrary commands when opened, and the child process received the developer's LaunchDarkly session token via LD_ACCESS_TOKEN. Affects Linux, macOS ($(...)) and Windows (" + &)

Fix: invoke `ld-find-code-refsrgument array. No shell is
involved, so shell metacharacters in any argument are inert.

// before
exec(`${bin} --dir="${directory}" --projKey="${project}" ...`, options)
// after
execFile(bin, ['--dir', directory, '--projKey', project, ...], options)

IFlagAliases.exec in src/models.ts gets the same signature. --contextLines=-1 stays in = form
so the negative value is not parse

Release

Version bumped to 6.2.2 with a Security entry in the CHANGELOG.

Verification

- tsc, eslint and prettier pass.


<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Overview**
> Fixes a **command injection** vulnerability in Code References alias generation by stopping use of shell-based `child_process.exec` when invoking `ld-find-code-refs`.
> 
> `FlagAliases` now uses **`execFile`** with a discrete argument array for `generateCsv` and `codeRefsVersionCheck`, so workspace-controlled values (`launchdarkly.project`, folder paths, etc.) are passed as argv and cannot inject shell metacharacters or exfiltrate `LD_ACCESS_TOKEN` from the child environment. **`IFlagAliases.exec`** is updated to match the new `(file, args, options)` signature.
> 
> Release **6.2.2** documents the security fix in the CHANGELOG.
> 
> <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit f356deb8551adc9b8289efb72a56f7ae3a78d258. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 6bbadcf. Configure here.}

[cursor]

Trust handler gated on version check

Medium Severity

The onDidGrantWorkspaceTrust listener is nested under codeRefsVersionCheck(), which returns false when the workspace is untrusted. That call is not awaited, so the Promise is always truthy and the listener still registers. Awaiting the check would skip this recovery and show a false version error instead.

 

^{Reviewed by Cursor Bugbot for commit 6bbadcf. Configure here.}

[cursor]

Trust listener registered multiple times

Low Severity

setupComponents can run again while the workspace is still untrusted, for example after a session change or configure. Each run pushes another onDidGrantWorkspaceTrust listener. Granting trust then calls start() multiple times on the same FlagAliases instance and starts overlapping refresh intervals.

 

^{Reviewed by Cursor Bugbot for commit 6bbadcf. Configure here.}