Using Spwnr

Use this skill as the main entry router for the Spwnr Claude Code plugin.

This plugin is a controller, not the worker itself.

Default workflow:

1. /spwnr-plan
2. /spwnr-task
3. implement

Main Route

• Start with /spwnr-plan for non-trivial work that needs a locked executable plan before implementation.

• /spwnr-plan should load spwnr-principle plus spwnr-plan, inspect context with Read, ask only material follow-up questions with AskUserQuestion, keep blockers visible with TodoWrite, and persist the latest active revision under .claude/plans/spwnr-<project-folder-name>-<YYYY-MM-DD>.md or .claude/plans/spwnr-<project-folder-name>-<YYYY-MM-DD>-rN.md.

• Planning should run a sequential expert loop with spwnr resolve-workers plus planning-only Agent passes for research, draft, and review.

• Planning must choose pipeline or team, explain why, and persist the execution pattern.

• After each write or revision, /spwnr-plan should run the execution review loop with AskUserQuestion, using Execute current plan, Continue improving plan, and End this round.

• Execute current plan is the handoff signal into /spwnr-task.

• Use /spwnr-task after the task already has a ready active revision.

• /spwnr-task should read the latest active revision with Read, append Approved Execution Spec with Edit, resolve registry candidates with resolve-workers, build per-unit coverage with repeatable --unit "<unit-id>::<brief>" inputs when needed, and route implementation through the helper docs in skills/spwnr-task/task-pipeline.md or skills/spwnr-task/task-team.md.

• If registry resolution cannot satisfy the approved capability requirements, stop and direct the user to /spwnr-worker-audit before retrying /spwnr-task.

Alternate Lanes

• Use /spwnr-do for a bounded small task that can be handled directly in the current run without the full plan artifact flow.

• /spwnr-do writes a lightweight runtime note under .claude/do/spwnr-do-<project-folder-name>-<YYYY-MM-DD-HHMMSS>-<slug>.md.

• /spwnr-do may use Read, Write, Edit, Agent, and spwnr resolve-workers, but it must never create tasks or teams.

• /spwnr-do may invoke at most 3 direct workers.

• If /spwnr-do finds a worker gap, redirect to /spwnr-worker-audit.

• If /spwnr-do discovers the task is broad, multi-stage, risky, or planning-sensitive, redirect to /spwnr-plan.

• Use /spwnr-worker-audit when registry health, local package availability, injected agent visibility, or readiness recovery is unclear.

• /spwnr-worker-audit is the deeper audit and recovery lane for install, inject, and sync guidance.

Quick Examples

• Broad coding or research request with several execution units -> /spwnr-plan, then /spwnr-task, then implement
• Approved plan already exists and the user wants execution now -> /spwnr-task
• Small follow-up fix after a completed workflow -> /spwnr-do
• Missing local packages, injection, or registry readiness -> /spwnr-worker-audit