How do I create an OpenCode status checkpoint before retained Montage work?
OpenCode’s remote OAuth defaults do not authorize every tool exposed after sign-in. Put a narrow checkpoint between registration and a larger agent task so that a saved server entry cannot silently become approval for retained research operations. Establish the checkpoint before letting the agent resume a larger brief. Remote OAuth defaults do not determine approval for a proposed write or make registration evidence equivalent to an executed workspace read.
Help me create an OpenCode status checkpoint before retained Montage work. Name the current purpose as verifying the intended Montage workspace. Record status-only success or the exact blocker and await a new instruction before research. No writes.Pause the broader objective
Name the current purpose as verifying the intended Montage workspace. Leave watchlist creation, audience changes and other writes outside this diagnostic regardless of the original coding task. A separate preliminary brief is important when the agent’s prior objective contains effects beyond reading status, even if those effects were discussed earlier.
Confirm the selected entry
Inspect the active v2 remote server registration and the session using it. Another project’s successful sign-in should not substitute for the current session’s own status evidence. The active session’s selected entry is the target of the check; another project’s working remote server is only comparative configuration evidence.
Request a single read
Ask for signals_status and review any tool approval prompt. If OpenCode proposes a different operation to test connectivity, stop rather than approving the expanded remote effect. Tool approval should be evaluated against the named operation, not justified by the fact that remote servers default to OAuth authentication.
Read the actual result
Compare workspace and coverage with expectations and preserve any authorization error. The presence of a server in opencode mcp list or /mcps is weaker evidence than this execution response. If the actual workspace differs, pause before private research instead of asking the agent to explore records to guess which account was selected.
Conclude with limited authority
Record status-only success or the exact blocker and await a new instruction before research. A configured remote OAuth server does not enlarge account allowances or grant blanket permission for future agent writes. Keep future task approval separate from successful login so a shared application session does not imply standing authorization for all discovered remote effects.
Keep the boundaries clear.
- Tool approval should be evaluated against the named operation, not justified by the fact that remote servers default to OAuth authentication.
- Keep future task approval separate from successful login so a shared application session does not imply standing authorization for all discovered remote effects.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.