Documentation Index
Fetch the complete documentation index at: https://mintlify.com/xantorres/repokernel/llms.txt
Use this file to discover all available pages before exploring further.
The run orchestrator automates the sprint loop for an epic: it queues, starts, hands off to an agent, waits for review, and closes sprints in sequence (or in parallel waves). rk run is the primary entry point for both long-running epic automation and one-shot fastpath tasks. The supporting commands — rk runs, rk run inspect, rk run logs, and rk run abort — give you visibility into and control over active and historical runs.
rk run <epic-id>
Start or resume an automated run loop for an epic.
rk run <epic-id> [--agent <name>] [--mode assisted|autonomous]
[--limit N] [--resume RUN-NNN] [--dry-run]
[--parallel] [--sequential] [--concurrency N]
[--lane <name>] [--allow-overlap]
[--worktree] [--cwd <path>]
| Flag | Default | Description |
|---|
--agent | manual | Agent to invoke: fake, manual, claude, codex, ollama, or a config-defined name |
--mode | assisted | assisted pauses for human review after each sprint; autonomous requires allowAutonomousClose: true in config |
--limit N | unlimited | Stop after N sprints complete |
--resume RUN-NNN | — | Resume a previously paused run by its run id |
--dry-run | — | Preview the run plan without executing any sprint |
--parallel | — | Assert parallel execution (the epic must declare execution_strategy: parallel) |
--sequential | — | Force sequential execution even on a parallel epic |
--concurrency N | config | Maximum sprints per wave during parallel execution |
--lane <name> | config default | Override the target lane for this run |
--allow-overlap | — | Allow overlapping allowed_paths across parallel sprints (requires parallel.allowOverlapFlag: true in config) |
--worktree | — | Force worktree isolation for sequential runs |
Fastpath: rk run -m "..."
Create a one-shot fastpath task (T-NNN) and immediately run it, without creating an epic or sprint manually:
rk run -m "Add /health endpoint" --agent claude
rk run -m "Fix null pointer in parser" --from-tracker gh:owner/repo#42 --agent claude
--from-tracker flag seeds the task title and body from an external issue (gh:owner/repo#NNN, jira:KEY-NN, linear:ABC-NN). The resulting task alias is available via rk task list and rk task status.
rk runs
List all run records for the project, with optional filtering.
rk runs [--status running|paused|completed|failed|aborted] [--epic <epic-id>] [--json] [--cwd <path>]
| Flag | Description |
|---|
--status | Filter to runs in a specific state: running, paused, completed, failed, or aborted |
--epic <epic-id> | Show only runs associated with the given epic |
--json | Machine-stable JSON output |
rk run inspect <run-id>
Show the full state of a run and its next steps.
rk run inspect RUN-001 [--json] [--cwd <path>]
rk run logs <run-id> [sprint-id]
Show logs for a run, or for a specific sprint within a run.
rk run logs RUN-001
rk run logs RUN-001 S-002
rk run abort <run-id>
Abort a paused run. Removes the run record. Sprint statuses are not changed — any in-progress or paused sprints retain their current state.
rk run abort RUN-001 [--cwd <path>]
rk run abort only removes the run orchestration record. To roll back work in a sprint, use rk reopen S-NNN after aborting.
Fastpath Task Commands
Fastpath tasks are T-NNN aliases synthesized by rk run -m "...". Mutation lives in rk run, rk close, and rk discard — there is no rk task close alias.
rk close T-001
Review, run checks, and merge the fastpath task to main.
rk discard T-001
Drop a fastpath task without merging.
rk task list
List every task alias the project has produced, sorted by id.
rk task list [--status active|review|shipped|cancelled] [--json] [--cwd <path>]
| Flag | Description |
|---|
--status | Filter to active, review, shipped, or cancelled |
--json | Emit the raw TaskAlias[] array; no envelope wrapper |
rk task status <id>
Show one task’s status, sprint linkage, source, created/closed timestamps, and (when present) a truncated review_sha.
rk task status T-001 [--json] [--cwd <path>]
--json emits the raw TaskAlias object. Accepts non-padded ids (T-1 resolves to T-001).
rk task inspect <id>
Same fields as rk task status plus the resolved on-disk paths to the alias JSON, the synthesized sprint markdown, and (when available) the review markdown.
rk task inspect T-001 [--json] [--cwd <path>]
--json emits an { alias, paths: { alias, sprint, review } } envelope. When the project graph fails to load, the command still surfaces the alias with a (not found) placeholder for the sprint and review paths rather than failing the entire call.