<objective> Analyze project state and route user to the appropriate agent. </objective> <triggers> - User runs `/specsmd-fire` on initialized project - After any agent completes its task </triggers> <llm critical="true"> <mandate>ALWAYS scan file system for intents/work-items not in state.yaml</mandate> <mandate>FILE SYSTEM is source of truth — state.yaml may be incomplete</mandate> <mandate>Route based on VERIFIED state, not assumptions</mandate> </llm> <flow> <step n="1" title="Discover and Read State"> <action>Read .specs-fire/state.yaml</action>

code

<file_system_scan critical="true">
  Use these EXACT glob patterns:

  <pattern purpose="Find intent briefs">
    .specs-fire/intents/*/brief.md
  </pattern>

  <pattern purpose="Find work items">
    .specs-fire/intents/*/work-items/*.md
  </pattern>

  Work items are {work-item-id}.md files directly in work-items/ folder.
</file_system_scan>

<action>Reconcile: add discovered items to state as pending</action>
<action>Parse current project state</action>

</step> <step n="2" title="Check Active Run"> <check if="runs.active is not empty"> <output> Resuming active run: {runs.active[0].id} Scope: {runs.active[0].scope} Current item: {runs.active[0].current_item} Progress: {completed_count}/{total_count} items </output> <route_to>builder-agent (run-execute)</route_to> <stop/> </check> </step> <step n="3" title="Check Pending Work Items"> <action>Find work items with status == pending across all intents</action> <check if="pending work items exist"> <output> **{pending_count} pending work items** found across {intent_count} intent(s).

code

    Plan run scope and start execution? [Y/n]
  </output>
  <check if="response == y">
    <route_to>builder-agent (run-plan)</route_to>
  </check>
  <stop/>
</check>

</step> <step n="4" title="Check Active Intent"> <action>Find intents with status == in_progress</action> <check if="active intent has no work items"> <output> Intent "{intent.title}" needs decomposition. Routing to Planner to create work items. </output> <route_to>planner-agent (work-item-decompose)</route_to> <stop/> </check> <check if="all work items completed"> <action>Mark intent as completed</action> <output> Intent "{intent.title}" completed!

code

    Work items delivered:
    {list completed work items}

    Ready for next intent? [Y/n]
  </output>
</check>

</step> <step n="5" title="No Active Work"> <output> No active work. Ready for a new intent.

code

  What do you want to build?
</output>
<route_to>planner-agent (intent-capture)</route_to>

</step> </flow>

<routing_decision_tree>

code

state.yaml + file system scan
    │
    ├── runs.active? ─────────────> Builder (run-execute, resume)
    │
    ├── pending work items? ──────> Builder (run-plan, then execute)
    │
    ├── intent without work items? > Planner (work-item-decompose)
    │
    └── no active intents ────────> Planner (intent-capture)

</routing_decision_tree>

<context_passed_to_agents> To Planner:

yaml

context:
  action: intent-capture | work-item-decompose
  intent_id: {if decomposing}

To Builder:

yaml

context:
  action: run-plan | run-execute | resume
  pending_items: [{list of pending work items}]  # for run-plan
  run_id: {if resuming}

</context_passed_to_agents>

<success_criteria> <criterion>File system scanned for untracked intents/work-items</criterion> <criterion>State reconciled with file system</criterion> <criterion>Correct agent selected based on state</criterion> <criterion>Context passed to target agent</criterion> </success_criteria>