Declare output schemas on polyphony's script-agent steps (conductor 9d603a1)

[PolyphonyRequiem]

Summary

conductor v0.1.17+ allows script: agents to declare \output:\ schemas just like LLM agents (#118 / commit 9d603a1 in microsoft/conductor). Polyphony's lifecycle-dispatch scripts (lifecycle-router.ps1, batch-integrator.ps1, etc.) currently emit JSON to stdout with no declared schema, so conductor cannot validate output fields at schema-parse time.

What to do

For each script node in .conductor/registry/workflows/\ that emits structured JSON, add an \output:\ schema declaration matching the emitted shape. This closes the M2-class footgun for script-node outputs and enables the Jinja path resolver lint to validate script outputs.

Files to audit

• 
  oot-item-dispatch.yaml\ — lifecycle-router, state-detect, emitter scripts
• 
  oot-batch-dispatch.yaml\ — aggregate_renegotiation, batch-integrator
• \polyphony.yaml\ — outer_loop_evaluator, preflight_sync
• Any other workflows with \ ype: script\ nodes that emit JSON

Effort/Risk

~1-2h per workflow (mechanical annotation pass). Low risk — adds schema; doesn't change runtime behavior.

References

• Conductor commit 9d603a1: \eat(script): allow script agents to declare output schemas\
• conductor-mechanics skill M2: LLM Agent Output Schemas (same footgun for script nodes)
• Mahler adoption survey 2026-05-28 (decision inbox)

/cc Epic #521