fix(cursor): honor sessionId in init for multi-turn resume (MT-247)

[ivanzzeth]

Summary

• Resolve Cursor CLI system/init session id the same way as the Claude adapter: accept session_id or sessionId, then fall back to the adapter-generated placeholder.
• Ensures continuation turns pass --resume with the real CLI session when only camelCase is present, so multi-turn AgentRunner runs stay on the same thread while a Linear issue remains in an active working state.

Test plan

• mise exec -- mix test test/symphony_elixir/cursor_adapter_test.exs:39 test/symphony_elixir/cursor_adapter_test.exs:63
• mise exec -- mix test test/symphony_elixir/core_test.exs:1230 (agent continuation with fake codex)

Refs: MT-247

Made with Cursor

[ivanzzeth]

Review: PR #58 — MT-247 cursor session resume

Summary

3-line change to handle camelCase sessionId in cursor stream-json init payload, plus test fixture.

Analysis

• Fallback chain session_id → sessionId → session.session_id correctly maintains backward compatibility
• Test script camel_session_init_script produces sessionId (camelCase) — verifies the parsing path

Note

PR #61 (fix/mt-247-agent-handoff-states) includes the exact same cursor adapter change plus the additional handoff fix. One of these should be closed to avoid merge conflicts.

Recommendation: Approve — but consider whether to close this in favor of PR #61 which contains both fixes for MT-247.

[ivanzzeth]

Review: PR #58 — MT-247 cursor session resume

Summary

3-line change to handle camelCase sessionId in cursor stream-json init payload, plus test fixture.

Analysis

• Fallback chain session_id → sessionId → session.session_id correctly maintains backward compatibility
• Test script camel_session_init_script produces sessionId (camelCase) — verifies the parsing path

Note

PR #61 (fix/mt-247-agent-handoff-states) includes the exact same cursor adapter change plus the additional handoff fix for the same Linear issue. One of these PRs should be closed to avoid duplicate commits.

Recommendation: Approve — but check if PR #61 supersedes this one for the same MT-247 issue.

[ivanzzeth]

Superseded by PR #61 which includes the same fix plus the handoff-state handling for MT-247.