Use when navigating Valdr state or routing work through orchestrator agents, including general PM discovery, registry operations, and sprint orchestration.
You are an AI agent routing Valdr work through orchestrator agents. This skill is a thin wrapper that loads the resolved orchestrator agent prompt and defers all orchestration details to it.
Tool naming: Code blocks use fully-qualified MCP tool names (mcp__valdr__pm_*). Short names may appear in prose for readability.
Before loading prompt content, determine which orchestrator should handle the request.
Validate the requested handle directly:
mcp__valdr__pm_agent → { action: "get", handle: "<orchestrator handle>" }
If the handle exists, use it. If it does not exist, STOP and ask the user for direction.
List the available orchestrators first:
mcp__valdr__pm_agent → { action: "list", defaultRoles: ["orchestrator"] }
| Result | Action |
|---|---|
| 1 or more orchestrators found | Present the available handles to the user and ask which orchestrator should handle the request |
| No orchestrators found | STOP: "No orchestrator agents found. Ask the user to register one in the registry." |
Load the resolved orchestrator agent prompt by handle:
mcp__valdr__pm_agent → { action: "get_prompt", handle: "<orchestrator handle>" }
Example after selection:
mcp__valdr__pm_agent → { action: "get_prompt", handle: "gunnar" }
If the prompt cannot be loaded, STOP and ask the user for direction.
Follow the loaded orchestrator prompt as the single source of truth. Hot-load any workflows or tool docs it references as needed.
| Task Type | Skill | Description |
|---|---|---|
| Task execution | valdr-executor | Running, implementing, or completing tasks |
| Feature planning | valdr-planner | VMP markdown plans, specs, requirements |
| Code reviews | valdr-reviewer | Review workflow, lightweight review scores, verification |
| Session auditing | valdr-auditor | Audit workflow, score runs, and execution quality |