How this skill is triggered — by the user, by Claude, or both
Slash command
/spwnr:spwnr-principleThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill as the shared rulebook for Spwnr planning, approval-gated execution, registry-guided worker selection, and runtime guardrails.
Use this skill as the shared rulebook for Spwnr planning, approval-gated execution, registry-guided worker selection, and runtime guardrails.
Skill before acting.AskUserQuestion for material decisions with 2 to 4 concrete options.TodoWrite for blockers, readiness fields, and approval condition.Write or Edit./spwnr-plan, a planning-only Agent pass is allowed only after a draft plan is visible.TaskCreate, TaskGet, TaskList, TaskUpdate, TeamCreate, TeamDelete, or SendMessage for non-trivial work until a draft plan exists and the current run has approved execution..claude/plans/spwnr-<project-folder-name>-<YYYY-MM-DD-HHMMSS>-<GOAL>.md-r2, -r3, and so on.Revision, Revision Status, Supersedes, and Superseded By.superseded; only the latest active revision is executable.A ready plan must lock goal, success criteria, scope boundaries, constraints, assumptions, risks, approval condition, active artifact path, execution strategy, executable units, capability requirements, failure rules, and ownership boundaries.
It must be explicit enough for a fresh worker to execute from the artifact without reconstructing intent from chat.
pipeline or team.Execution Strategy Recommendation.pipeline, persist pattern name, ordered stages, stage objectives, inputs, outputs, acceptance checks, capability tags, and handoffs.team, state whether to launch multiple bounded pipelines or one shared queue.pipeline must work without Claude team features.Each Execution Unit must include:
unit_idobjectivepreconditionsinputsimplementation stepsexpected outputacceptance checkdependenciespreferred capability tagsrisk levelfile ownership hintsclaim policy recommendationheartbeat expectationworker plan approvalpipeline pattern reference or overrideRender field labels as single-line bullets. Preferred marker: - **unit_id**: unit-1. Compatible fallbacks include - **unit_id:** unit-1, - unit_id: unit-1, and full-width colon variants; do not hide unit_id in tables or prose.
Every execution, integration, and review task created with TaskCreate must include:
Plan: <path>Unit: <unit-id>Mode: <pipeline|team>Worktree: <required|optional|not-required>Blocked: noOwner: <agent-name|controller|unassigned>Files: <csv scope or none>Claim-Policy: <assigned|self-claim>Risk: <low|medium|high>Plan-Approval: <not-required|required|approved>The referenced plan must contain Approved Execution Spec. Blocked: is reserved for current block state only. New tasks write Blocked: no; sequencing belongs in plan dependencies, graph edges, or Depends-On:.
Claim-Policy: assigned requires a concrete Owner.Claim-Policy: self-claim starts with Owner: unassigned.Risk: high requires Plan-Approval: required or approved.When the launch policy targets Claude Code with worktree-required mutation isolation, Claude mutating tasks default to Worktree: required; reserve Worktree: not-required for read-only review/audit. Mutating work must discover ToolSearchTool, enter with EnterWorktreeTool, summarize with BriefTool, and exit with ExitWorktreeTool.
Before the first TaskCreate, check marker presence, compatibility matrix, Blocked: no, disjoint team file ownership, shared-file exceptions, worktree policy, and high-risk approval state.
After every plan write or revision, ask:
Execute current plan — only execution permission signal; hand off to /spwnr-taskContinue improving plan — revise the same active revision unless this becomes a material re-planEnd this round — preserve the artifact and stop cleanlyRecord review history in the artifact; do not recreate a persistent approval state machine.
High-risk units require Risk: high, Plan-Approval: required, a teammate mini-plan, controller approval, then Plan-Approval: approved before mutation.
When registry or capability gaps block progress, stop with:
Plan ArtifactReadiness GapMissing <Capability Type>Recovery Steps — direct to /spwnr-worker-audit, preserve the active revision, and never silently downgradeNext StepGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub cheong43/spwnr --plugin spwnr