Rote Command Patterns

rote-command-patterns

All rote-<name> references in this document — including every name in the Handoff Contract — are companion skills, never CLI commands (rote-shell is not rote shell). Invoke them through the runtime's skill mechanism; only literal rote … commands run in a terminal.

Use this skill when command syntax or rote idioms are needed for an active rote task and live help is not enough by itself. Prefer live surfaces first: rote start, rote guidance <area> essential, and rote grammar <topic>.

This skill does not own task execution. It returns command patterns to the skill that already owns the workflow.

Live Surfaces First

• Use rote grammar query for cached response addressing and jq flags.
• Use rote grammar adapters and rote guidance adapters essential before adapter probes, calls, sessions, or batches.
• Use rote grammar deno and rote guidance typescript essential before TypeScript play execution.
• Use rote grammar export before play scaffolding, pending play work, lint, release, or index commands.
• Use rote grammar registry before pushes, pulls, namespace selection, or publish steps.

If live grammar conflicts with this skill, live grammar wins.

Adapter Probe And Call

Start in rote-workspace, then inspect before calling:

rote <adapter_id>_probe "<operation intent>"
rote <adapter_id>_call <operation> '<args-json>'

Convert hyphenated adapter IDs to underscores in the command name. Read the full probe response, copy the exact operation name and required JSON shape, then call through rote. Read @@next, @@mandatory, @@plays, warnings, and cached response ids before continuing.

Response Querying

Use cached response IDs instead of copying JSON:

rote query @1 '<jq-filter>' -r

Quote jq filters with single quotes unless the filter intentionally needs shell variables. Use -r only when raw strings are wanted. If a query fails, inspect whether the cached response is an error envelope before changing jq syntax.

Prefer rote-native, POSIX-friendly response methods before raw filesystem access:

• Use rote query @N '<jq-filter>' for extraction and rote query is-error @N for error checks.
• Use rote grammar query when syntax, raw output flags, or response addressing are unclear.
• Use rote-provided Deno/TypeScript transformations for complex reductions that do not fit jq.
• Keep transformations attached to response IDs so provenance remains visible in the workspace.

Do not read .rote/responses/@N.json directly for normal analysis, and do not pipe those internal files through ad hoc scripts. Direct response-file inspection is a troubleshooting exception only: use it after rote-native query or state inspection cannot recover the needed data, record why, and return to rote commands as soon as the cause is known.

Prefer separate adapter/workspace lifecycle commands for setup, repair, pending save, release, index, and registry verification so each output is read before the next step. Captured shell command chains belong under rote-shell; use them only when each meaningful process is run through rote proc.

Batch Calls And Queues

Use batch patterns only after a single call proves the operation and input shape. Keep the batch in rote so responses remain cached and queryable. For long-running or queued work, capture the task ID, poll through rote commands, and query cached results when complete instead of managing background processes directly.

Browser And Registry Handoffs

For browser automation, hand off to rote-browse and use rote guidance browser essential for live browser conventions. Keep browser-derived data flowing through rote workspace state before crystallizing a reusable play.

For registry publishing, use live registry guidance and grammar, confirm the target namespace and artifact, then hand off to rote-registry. If organization administration is needed, hand off to rote-org through rote-registry.

Return Fields

Return these fields to the owning skill:

• Live grammar or guidance command checked.
• Pattern selected and exact command shape.
• Quoting, cwd, workspace, or response-id caveats.
• Owning skill to resume: rote-workspace, rote-flow-authoring, rote-registry, rote-browse, or rote.

Handoff Contract

• Use when: an active rote workflow needs command idioms and live grammar/guidance alone is insufficient or needs task-specific interpretation.
• Preconditions: another skill owns the task branch, and the runtime can run rote grammar or report why live command help is unavailable.
• Owns: command-pattern lookup, live-surface precedence, quoting/cwd caveats, and return of exact command shapes to the owning skill.
• Hands off to: rote-workspace for adapter execution, rote-flow-authoring for play lifecycle commands, rote-typescript-transformations for TypeScript execution detail, rote-registry for publish/pull commands, rote-browse for browser commands, and rote-troubleshooting for repeated failures.
• Returns to: the skill that requested command guidance with the selected pattern and live source of truth.
• Stop when: the owning skill has enough syntax to continue, live grammar supersedes this guidance, or repeated failures require troubleshooting.
• Completion signal: exact command pattern, caveats, and owning skill to resume are named.