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.
Review and gate commands form the quality control layer between sprint work and merge. rk review opens a review stub, rk review-verdict records a human or automated decision, and rk review-evidence appends command output as proof. The rk gates command runs the full check bundle for a sprint; rk gate manages human checkpoints that pause autonomous runs. rk reject records an ADR for work explicitly ruled out.
rk review <sprint-id>
Create a review stub and transition the sprint to review.
rk review S-001 [--cwd <path>]
automation.reviewer for its reviewer: frontmatter value, falling back to automation.defaultReviewer. The review file is staged but not committed — the agent or human reviewer modifies it before calling rk review-verdict.
rk review-verdict <review-id> <verdict>
Set the verdict on a review entity.
rk review-verdict R-001 accepted [--summary "..."] [--cwd <path>]
rk review-verdict R-001 changes_requested
rk review-verdict R-001 rejected
| Flag | Description |
|---|
--summary "..." | Optional human-readable summary written to review frontmatter |
Verdicts
| Value | Meaning |
|---|
accepted | Work meets acceptance criteria; sprint is eligible for rk close |
changes_requested | Reviewer has requested specific changes before acceptance |
rejected | Work is declined; sprint should be reopened or cancelled |
rk review-verdict writes the verdict to frontmatter and stages the file. It does not commit. After setting a verdict, resume the paused orchestrator run with rk run --resume RUN-NNN.
Exit 0 on success, 1 if the review id is not found or already terminal, 2 on runtime error.
rk review-evidence <id>
Append command evidence to a review. The target may be a review id or a sprint id with a linked review.
rk review-evidence S-001 --label focused-tests --command "pnpm test -- filter" [--summary "..."] [--json]
rk review-evidence R-001 --label full-gates --command "rk gates S-001"
command_evidence[]. rk ship and rk gates record their own evidence automatically.
The --exit-code flag imports already-run evidence without executing the command — this is for importing external results and does not satisfy review gates.
rk gates <sprint-id>
Run the repo-configured gate bundle for a sprint. The bundle is repo-agnostic and never hardcodes a package manager.
rk gates S-001 [--json] [--cwd <path>]
automation.checksCmd (when configured in repokernel.config.yaml)- Diff and path checks against
allowed_paths / denied_paths
rk validate --fail-on P1rk registry --check --explain
The output prints allowed_paths and denied_paths before the risky checks so the scope is visible upfront. If the sprint has a linked review, the command appends each check as command_evidence.
rk gate <subcommand>
Human checkpoints that pause an autonomous run before a specific sprint. Distinct from rk gates (the per-sprint check bundle above).
rk gate ls [--epic <id>] [--json] [--cwd <path>]
rk gate add <gate-name> --sprint <S-NNN> [--json] [--cwd <path>]
rk gate resolve <gate-name> [--epic <id>] [--force] [--dry-run] [--cwd <path>]
rk gate ls
List all declared gates, with optional filtering to an epic.
rk gate ls [--epic E-001] [--json] [--cwd <path>]
rk gate add
Declare a named gate on one or more planned, pending, or queued sprints. When the orchestrator reaches a gated sprint it pauses and prints instructions for resolution.
rk gate add security-review --sprint S-005 [--json] [--cwd <path>]
--sprint is repeatable. A sprint that has already started is rejected.
rk gate resolve
Clear a gate so the paused run can continue.
rk gate resolve security-review [--epic E-001] [--force] [--dry-run] [--cwd <path>]
| Flag | Description |
|---|
--epic <id> | Scope the resolution to gates on a specific epic |
--force | Resolve even if the gate has not been explicitly satisfied |
--dry-run | Preview the resolution without writing |
rk reject
Record an append-only rejection ADR for work the project has explicitly ruled out. The pattern is compiled as a JavaScript regex and matched against tracker title and body during future intake.
rk reject --pattern <regex> --reason <text> --scope feature|bug|enhancement \
[--ref <source>:<ref>] [--close] [--json] [--cwd <path>]
| Flag | Description |
|---|
--pattern <regex> | Non-empty JavaScript regex. Malformed or unsafe patterns exit EXIT_USAGE (64) before any disk write. |
--reason <text> | Human-readable rationale, minimum 20 characters |
--scope <scope> | Rejection category: feature, bug, or enhancement |
--ref <ref> | Optional source issue (gh:owner/repo#NNN, jira:KEY-NN, linear:ABC-NN). Malformed refs exit EXIT_USAGE (64). |
--close | Attempt a tracker comment plus close transition. Requires --ref. Tracker write failures return EXIT_RUNTIME (2) while preserving the local ADR. |
--json | Machine-stable JSON output with ok, action, id, and tracker status |
(pattern, scope) writes are idempotent and return the existing ADR rather than creating a duplicate.