feat(agent-relay-cursor): add share_desktop

[matifali]

Description

Adds share_desktop to agent-relay-cursor. When set, the worker starts with --share-desktop=view_and_control, so people can watch or take control of the agent's desktop from the Desktop tab in Cursor. It requires computer_use = true (validated) and TigerVNC/Xfce in the image, e.g. ghcr.io/coder/agent-relay-workspace:desktop (coder/agent-relay#86).

The mode is passed explicitly because --share-desktop takes an optional argument and, bare, swallows the following start.

Type of Change

• Feature/enhancement

Module Information

Path: registry/coder/modules/agent-relay-cursor
New version: v0.4.0
Breaking change: [ ] Yes [x] No

Testing & Validation

• Tests pass (bun test: 19 pass; terraform test: 12 pass)
• Code formatted (bun fmt)
• Changes tested locally: end to end on a local Coder with Agent Relay, a Cursor pool agent was dispatched into a workspace on the desktop image with computer_use + share_desktop. The worker ran with --computer-use --share-desktop=view_and_control, started Xtigervnc and Xfce, and agent worker debug reported computer use ready and desktop share supported. The agent's desktop showed up live in Cursor's Desktop tab.

Related Issues

coder/agent-relay#86

https://claude.ai/code/session_01L2wLoUuHsK2Nzm93TjjoMi

[github-actions]

Module Scorecard Check

coder/agent-relay-cursor: 58 → 58

✅ Score unchanged at 58 / 100. This PR does not affect the module's scorecard; the results are still good.

Theme	Before	After
Presentation & Onboarding	6 / 17	6 / 17
Integration	5 / 25	5 / 25
Credential Hygiene	20 / 20	20 / 20
Restricted-Environment	12 / 20	12 / 20
Engineering Quality	10 / 10	10 / 10
Overall	58 / 100	58 / 100

Full scorecard for this PR

Presentation & Onboarding	Agent Integration	Credential Hygiene	Restricted-Environment Readiness	Engineering Quality	Overall
6 / 17	5 / 25	20 / 20	12 / 20	10 / 10	58 / 100

Drilldown

Presentation & Onboarding — 6 / 17

Criterion	Max	Score	Notes
Configuration-mode examples	12	6	The README shows one full example (default install_cli = true). Other major modes—install_cli = false (baked image), computer_use = true, share_desktop = true—are mentioned in the Requirements section but lack dedicated examples with sensible defaults. Partial credit for documenting the primary mode and referencing the others.
Visual preview	5	0	No image, GIF, or video is embedded in the README. The frontmatter references an SVG icon (icon: ../../../../.icons/cursor.svg), but icons do not count.

Credential Hygiene — 20 / 20

Criterion	Max	Score	Notes
Secrets marked sensitive	16	16	agent_relay_credential is declared with mask_input = true and ephemeral = true in its styling block. The README example contains no inline secrets—only agent_id and install_cli. The "Worker credential" section explicitly states the token "is never written to disk."
Non-hardcoded auth path	4	4	The "Worker credential" section documents that Agent Relay mints a per-user sub-token at dispatch time; the service-account key never leaves Agent Relay. The README shows the full flow (relay exchanges key → stamps ephemeral parameter → module reads from environment). No raw keys are pasted into templates.

Restricted-Environment Readiness — 12 / 20

Criterion	Max	Score	Notes
Mirrorable artifact source	5	0	The install URL https://cursor.com/install is hardcoded in install.sh.tftpl. No module input variable overrides this URL. cli_binary changes the binary path, not the download source. No variable exists to point the installer at an internal mirror.
Bring-your-own binary	10	10	install_cli = false is documented in the README ("Bake the CLI into the image and set this to false for faster workspaces"). The install script becomes a no-op, and the start script still checks ~/.local/bin and PATH for the binary. Fully documented and tested.
Egress transparency	3	0	No dedicated README section enumerates external endpoints. cursor.com/install appears in the variable description and install script; cursor.com/agents/<id> and cursor:// appear in the app descriptions. These are scattered across unrelated sections with no air-gapped or restricted-network guidance.
Runs without sudo	2	2	All three scripts (install.sh.tftpl, start.sh.tftpl, status.sh.tftpl) never invoke sudo. The installer writes to ~/.local/bin; the start script writes to $HOME/.coder-modules/.... No root required for core functionality.

Engineering Quality — 10 / 10

Criterion	Max	Score	Notes
Input quality	6	6	All 8 module variables have clear descriptions and sensible defaults. cli_binary has a regex validation rejecting shell metacharacters; share_desktop validates the computer_use dependency; serving_log_pattern rejects empty strings. The agent_relay_cursor_credential_kind parameter includes named options.
Test coverage	4	4	main.tftest.hcl (12 test runs) covers the parameter contract, script rendering, credential wiring, and validation failures. main.test.ts (14 test cases) exercises the actual install/start scripts in a container with stub binaries, verifying lifecycle states, restart idempotency, credential routing, shell-injection safety, and file layout. Clear separation: HCL tests cover Terraform-level logic; TS tests cover end-to-end script behavior.

Agent Integration — 5 / 25

Criterion	Max	Score	Notes
AI governance	10	0	No mention of Coder AI Gateway or Agent Firewall in the README or code. The credential system is Agent Relay's own token-minting flow, not AI Gateway routing or Agent Firewall policy enforcement.
Dashboard entry point	5	5	Two coder_app resources are declared: cursor_web ("Open in Cursor Web") and cursor_desktop ("Open in Cursor Desktop"), both conditional on agent_relay_session_id being set. Documented in the README with the exact URLs and behavior.
Session continuity	5	0	The README documents worker-process restart safety (re-running the start step leaves a live worker alone), but does not document resuming a Cursor chat session across reconnects or relaunches. No mention of session-ID resumption, tmux, screen, or boo. Session continuity is delegated to Cursor/Agent Relay, not the module.
Managed configuration	5	0	No documentation of managed MCP servers, agent settings, policies, or workdir configuration. The module runs the Cursor CLI worker but does not expose or document any managed-configuration surface.

Overall — 58 / 100

Raw 53 / 92 → round(53 / 92 × 100) = 58

Tip

You can run this locally by telling your agent: "review this module against .github/scorecard/SCORECARD.md".

---

Scored against SCORECARD.md with solstice-1. Language-model scores are advisory.