Scope
Make long-running and policy-affecting slash commands obey the same control-plane contract: inspect, pause/redirect/cancel when applicable, resume/recover when possible, and leave a durable command receipt.
Command families in scope:
- Agentic work:
/agent, /subagents, /rlm, /review
- Background execution:
/jobs, /hooks
- Recovery:
/undo, /retry, /restore
- Trust and containment:
/trust, /network, permission-policy related config
Acceptance
- Every long-running command has an inspectable state row and cancellation/recovery story.
- Subagent/RLM summarized output exposes the full transcript/detail handle when needed.
- Policy commands report exactly what changed, where it persisted, and how to revert.
- Recovery commands explain whether they changed files, conversation state, both, or neither.
Related: #874, #891, #1364, #1186, #1641, #1768.
Scope
Make long-running and policy-affecting slash commands obey the same control-plane contract: inspect, pause/redirect/cancel when applicable, resume/recover when possible, and leave a durable command receipt.
Command families in scope:
/agent,/subagents,/rlm,/review/jobs,/hooks/undo,/retry,/restore/trust,/network, permission-policy related configAcceptance
Related: #874, #891, #1364, #1186, #1641, #1768.