spwnr-worker-audit

Use for /spwnr-worker-audit. Inspect registry readiness and current Claude agent availability for general-task workflows.

Spwnr Worker Audit

Use this skill to audit whether the Spwnr registry can dynamically resolve a usable agent lineup from the local registry.

This skill is the shared source of truth for dynamic registry readiness plus install or inject guidance.

It is a health-check and recovery surface, not a prerequisite step for /spwnr-task, which performs its own registry resolution during normal execution.

Required Checks

  1. Check whether .claude-plugin/workers.json exists.
  2. If it exists, read .claude-plugin/workers.json with Read. If it does not exist, treat the project as using the default dynamic worker policy.
  3. Confirm the worker policy is in dynamic mode. Missing config still means dynamic mode is active by default.
  4. Check project-scoped agents under .claude/agents/.
  5. Check user-scoped agents under ~/.claude/agents/.
  6. Determine whether spwnr is available directly.
  7. If spwnr is not available directly, tell the user to install the published spwnr-cli package before continuing. Do not redirect them to local development commands.
  8. Check whether vendor/spwnr-registry exists in the current repository.
  9. Inspect the local Spwnr registry and determine whether published packages exist at all, and whether dynamic resolution is likely usable.

Dynamic Readiness Rules

  • Treat the local Spwnr registry as the runtime source of truth.
  • Treat a missing .claude-plugin/workers.json as the built-in default dynamic policy, not as an unconfigured or disabled state.
  • Do not assume vendored templates are usable until they have been synced into the local registry.
  • Report already-injected project and user agents as current availability, not as the selection source of truth.
  • If the vendored registry exists but the local registry looks sparse or empty, call out that sync-registry is likely needed.
  • If /spwnr-task reported a registry gap, frame this command as the required recovery step before the task should continue.
  • Do not silently invent a fallback agent lineup.

Reporting Format

Use these sections:

  1. Worker Policy
  2. Local Registry
  3. Injected Agents
  4. Readiness Gaps
  5. Install Suggestions
  6. Recommended Next Step

Install Suggestions

When the local registry is missing usable packages, prefer these exact commands:

  • sync vendored community templates into the local registry
    • spwnr sync-registry
  • preview registry candidates for a task
    • spwnr resolve-workers --search "<search query>" --host claude_code --format json

When a specific agent must be present immediately in Claude Code, prefer:

  • user scope
    • spwnr inject <name> --host claude_code --scope user
  • repo scope
    • spwnr inject <name> --host claude_code --scope project

Rules

  • Report whether dynamic selection is enabled and quote the key policy fields.
  • If .claude-plugin/workers.json is absent, explicitly say that dynamic selection is still enabled by default and note that the built-in defaults are registrySource=local, selectionMethod=llm_choose, missingPolicy=auto_install_local, and lineup 1-4.
  • Report whether the local registry has enough published packages to support a useful candidate pool.
  • Report whether vendored registry content exists but still needs syncing.
  • If all readiness conditions are satisfied, say that the registry audit path is healthy and remind the user that /spwnr-task can resolve its lineup directly.
  • If /spwnr-task is blocked, explicitly say that the user should install or inject the missing agents first, then return to the same active revision and rerun /spwnr-task.
  • If a task brief is explicitly provided, you may suggest using resolve-workers to preview likely candidates, but do not mutate state from this skill.
spwnr-worker-audit — skill by Cheong43 | Shared Context