openpencil/tests
Danila Poyarkov bd7acd6e34
fix(desktop): pin every command line the app may start (#922)
* fix(desktop): pin every command line the app may start

On Windows the app starts npm-installed CLIs through cmd /c, and the shell scope let cmd take any arguments, so any code running in the webview could run any command. Each program now has a scope entry with its exact arguments, and Windows shims go through their own cmd-<name> entries with fixed /c <name> arguments. The agents and the MCP server no longer accept arbitrary arguments either. A test checks that every command the app starts has a matching entry on both platforms.

* test(desktop): expect Windows shims through their own scope entries

* test(desktop): check the executable of each shell scope entry

* test(desktop): allow no shell scope entry beyond the programs the app starts

An extra entry with a permissive validator passed the per-program checks.
2026-10-06 11:19:05 +00:00
..
app refactor(fig): export the symbol readers and move the library tests home (#929) 2026-10-06 10:44:22 +00:00
e2e feat(collab)!: sync layer trees as a CRDT and open each room in its own tab (#902) 2026-10-06 10:40:25 +00:00
engine fix(desktop): pin every command line the app may start (#922) 2026-10-06 11:19:05 +00:00
figma
fixtures test(fig): pin override and variable precedence to what Figma renders (#912) 2026-10-06 09:08:51 +00:00
helpers refactor(fig): export the symbol readers and move the library tests home (#929) 2026-10-06 10:44:22 +00:00
integration test: typecheck the test suites and fix what that found (#896) 2026-10-05 12:42:38 +00:00
AGENTS.md test: typecheck the test suites and fix what that found (#896) 2026-10-05 12:42:38 +00:00