Problem
Supabase evals already support Claude Code, Codex, OpenCode, and the in-process AI SDK agent, but not Cursor’s CLI agent (cursor-agent). We want to run the same tools-mode / skills comparisons against Composer 2.5 and Grok 4.5.
Proposal
Add a Cursor CLI harness in the same shape as peer agents:
- Runner — pinned official install,
CURSOR_API_KEY, headless --print --output-format=stream-json (+ --yolo --trust --approve-mcps)
- Parser — stream-json → shared
TranscriptEvents
- Experiments — baseline + no-skills for both models (Supabase MCP on, matching Claude Code)
Model ids from live cursor-agent --list-models: composer-2.5, cursor-grok-4.5-high.
Acceptance
Related
Implementation PR: #166
Problem
Supabase evals already support Claude Code, Codex, OpenCode, and the in-process AI SDK agent, but not Cursor’s CLI agent (
cursor-agent). We want to run the same tools-mode / skills comparisons against Composer 2.5 and Grok 4.5.Proposal
Add a Cursor CLI harness in the same shape as peer agents:
CURSOR_API_KEY, headless--print --output-format=stream-json(+--yolo --trust --approve-mcps)TranscriptEventsModel ids from live
cursor-agent --list-models:composer-2.5,cursor-grok-4.5-high.Acceptance
cursorAgent({ model })CURSOR_API_KEYin.env.exampleRelated
Implementation PR: #166