cancel

/ouroboros:cancel

Cancel stuck or orphaned executions and background jobs through the Ouroboros MCP tools, which work on hosts without shell access.

/ouroboros:cancel                 # list this project's active executions, ask which to cancel
/ouroboros:cancel <id>            # cancel one execution (orch_... or execution id) or job (job_...)
/ouroboros:cancel --all           # cancel every running or paused execution of this project

Instructions

Loading MCP tools (deferred-schema guard)

Ouroboros tools may be deferred, and a schema loaded on one turn can unload by the next. Immediately before each ouroboros_* call, load it with the active runtime's tool-discovery capability: "+ouroboros project status" for ouroboros_project_status; "+ouroboros cancel" for ouroboros_cancel_execution and ouroboros_cancel_job. A call whose schema is not loaded fails with "Invalid tool parameters". If the tool is already callable, discovery is a no-op and an empty result is expected. Use the fallback only when the tool is genuinely absent.

Listing active executions

Call ouroboros_project_status with project_dir set to the absolute path of the current project. Active executions are the runs whose status is running or paused; show each one's session_id, started_at, and seed_goal. If the call fails because the project has more runs than limit, retry once with limit set to the reported run count. The listing covers only runs attributed to this project. Background jobs have no listing tool, so also offer any job_... id returned by a start tool earlier in this conversation.

Steps

• No ID, not "all": list as above, ask which to cancel with the active runtime's ask_user capability, then cancel the chosen ids as below. If nothing is active, say so (and, when run_shell is available, mention ouroboros cancel execution --list for executions outside this project). If the user declines, cancel nothing.
• ID given: an id starting with job_ is a background job: call ouroboros_cancel_job with job_id. Any other id (an orch_... session id or an execution id) is an execution: call ouroboros_cancel_execution with execution_id and, if the user gave one, reason.
• "cancel all" / "cancel everything": there is no bulk cancel tool. List as above and call ouroboros_cancel_execution once per running or paused run. Say this covered this project's runs only; for every execution on the machine, ouroboros cancel execution --all needs run_shell.

Reporting outcomes

• ouroboros_cancel_execution: new_status cancelled means it is cancelled. cancellation_requested means the live worker was asked to stop and will record the cancellation itself; suggest ooo status to confirm. An "already in terminal state" error means there was nothing to cancel; skip it with a warning. "Execution not found" means the id is wrong. "Held by another live owner" means only the process that owns it can cancel it; relay that verbatim. "Failed to persist cancellation" is retriable; retry once.
• ouroboros_cancel_job: status cancel_requested means the job and its linked execution were asked to stop; cancelled means done. Any other terminal status (completed, failed, interrupted) means the job had already finished and nothing was cancelled. "Job not found" means the id is wrong.

Relay any other error verbatim. Then suggest ooo status to verify or ooo run to start fresh.

Fallback (no MCP server)

When the cancel tools are genuinely absent and the active runtime has run_shell, use the CLI, which writes cancellation events straight to the event store: ouroboros cancel execution --list (prints active executions without prompting; then ask with ask_user), ouroboros cancel execution <session_id>, or ouroboros cancel execution --all. Add --reason "<text>" to record why, and report what it printed (Cancelled execution: <id>, Cancellation requested for execution: <id>, or No active executions found.) with any warning verbatim. Without run_shell, tell the user cancellation needs the Ouroboros MCP server (ooo setup) or a shell.

RFC #1392 State Breadcrumb Footer

End the final response with exactly one line ◆ <current state> → next: <recommended action>. Derive the state from live session state via ouroboros_session_status when available, otherwise from this skill's actual outcome. Never use a linear Step N of M footer; when the next step is a real choice, list 2-3 options.