Demo Readiness Team

/demo_readiness_team — Autonomous Demo Readiness Skill

Composes production sweeps into a continuous rehearsal loop with bounded authority. Operates autonomously between sessions: rehearse the demo path, detect drift vs a proof-of-green bundle, auto-fix low-risk issues (bounded by concurrency and cost limits), and produce a morning handoff plan for human review.

Authority boundary: this skill NEVER deploys runtime changes, restarts production lanes, merges topology-affecting PRs, or mutates production data. Background agents are advisory by default. Topology and runtime-affecting actions require explicit human approval.

Usage

/demo_readiness_team rehearse [run_id]
/demo_readiness_team detect-drift [run_id] [--proof-of-green path/to/bundle.json]
/demo_readiness_team dispatch-fixes [run_id] [--dry-run]
/demo_readiness_team morning-handoff [run_id]
/demo_readiness_team full-loop [run_id] [--dry-run]

Behavior

rehearse

Runs the full demo rehearsal pipeline via node_demo_rehearsal. Executes the demo command envelope, captures runtime topology manifest, checks projection row, probes dashboard API, takes screenshots, and writes rehearsal_bundle.json to docs/evidence/demo-readiness/<run_id>/.

• Emits: onex.cmd.omnimarket.demo-rehearsal-start.v1
• Artifact: docs/evidence/demo-readiness/<run_id>/rehearsal_bundle.json
• Status: GREEN / DEGRADED / BROKEN

detect-drift

Diffs current state vs an existing proof-of-green bundle via node_demo_drift_detector. Uses truth hierarchy: topology > projection > dashboard > screenshots.

• Emits: onex.cmd.omnimarket.demo-drift-detect-start.v1
• Artifact: docs/evidence/demo-readiness/<run_id>/drift_report.json
• Criticality levels: DEMO_BLOCKER, DEMO_DEGRADED, COSMETIC, OBSERVABILITY_ONLY, BACKLOG_ONLY

dispatch-fixes

Auto-fixes low-risk drift (COSMETIC, OBSERVABILITY_ONLY) via node_demo_fix_dispatcher. DEMO_BLOCKER and DEMO_DEGRADED findings require human approval. Respects bounded concurrency limits: max 4 parallel workers, max $50/day cost, max 5 open auto-fix PRs.

• Emits: onex.cmd.omnimarket.demo-fix-dispatch-start.v1
• Artifact: docs/evidence/demo-readiness/<run_id>/fix_dispatch_log.json

morning-handoff

Produces human-readable summary and machine-readable morning_dispatch_plan.json via node_morning_handoff_generator. Synthesizes overnight rehearsal runs and drift reports.

• Emits: onex.cmd.omnimarket.morning-handoff-start.v1
• Artifact: docs/evidence/demo-readiness/<run_id>/morning_dispatch_plan.json

full-loop

Runs rehearse → detect-drift → dispatch-fixes → morning-handoff as a single unattended loop. Use for overnight/cron automation. With --dry-run, skips dispatch-fixes.

Replay Mode

/demo_readiness_team morning-handoff <run_id> --replay

Loads saved evidence from docs/evidence/demo-readiness/<run_id>/ without re-running sweeps.

Bounded Authority

Concurrency Limits

max_parallel_workers: 4
max_runtime_minutes: 120
max_daily_cost_usd: 50.0
max_open_autofix_prs: 5

Evidence Directory

All artifacts written to docs/evidence/demo-readiness/<run_id>/:

• rehearsal_bundle.json — full rehearsal result with correlation data
• drift_report.json — drift findings with criticality classifications
• fix_dispatch_log.json — auto-fix dispatch log and results
• morning_dispatch_plan.json — human-readable + machine-readable handoff plan

Implementation

The skill delegates to four omnimarket nodes:

• node_demo_rehearsal — executes demo path, captures evidence bundle
• node_demo_drift_detector — diffs vs proof-of-green using truth hierarchy
• node_demo_fix_dispatcher — auto-fix dispatcher with bounded authority
• node_morning_handoff_generator — overnight summary + dispatch plan

All nodes are zero-infra capable (EventBusInmemory, filesystem-only artifacts).

rehearse invocation

import asyncio
from omnimarket.nodes.node_demo_rehearsal.handlers.handler_demo_rehearsal import (
    HandlerDemoRehearsal, ModelDemoRehearsalRequest,
)
result = asyncio.run(HandlerDemoRehearsal().handle(
    ModelDemoRehearsalRequest(
        run_id="<run_id>",
        dry_run=False,
    )
))
# result.rehearsal_bundle, result.bundle_path, result.overall_status

detect-drift invocation

import asyncio
from omnimarket.nodes.node_demo_drift_detector.handlers.handler_demo_drift_detector import (
    HandlerDemoDriftDetector, ModelDemoDriftDetectRequest,
)
result = asyncio.run(HandlerDemoDriftDetector().handle(
    ModelDemoDriftDetectRequest(
        run_id="<run_id>",
        proof_of_green_path="docs/evidence/demo-readiness/<green_run_id>/rehearsal_bundle.json",
    )
))
# result.drift_findings, result.demo_blocker_count, result.report_path

dispatch-fixes invocation

import asyncio
from omnimarket.nodes.node_demo_fix_dispatcher.handlers.handler_demo_fix_dispatcher import (
    HandlerDemoFixDispatcher, ModelDemoFixDispatchRequest,
)
result = asyncio.run(HandlerDemoFixDispatcher().handle(
    ModelDemoFixDispatchRequest(
        run_id="<run_id>",
        drift_report_path="docs/evidence/demo-readiness/<run_id>/drift_report.json",
        dry_run=False,
    )
))
# result.fixes_dispatched, result.fixes_skipped_human_approval, result.dispatch_log_path

morning-handoff invocation

import asyncio
from omnimarket.nodes.node_morning_handoff_generator.handlers.handler_morning_handoff_generator import (
    HandlerMorningHandoffGenerator, ModelMorningHandoffRequest,
)
result = asyncio.run(HandlerMorningHandoffGenerator().handle(
    ModelMorningHandoffRequest(
        run_id="<run_id>",
        evidence_dir="docs/evidence/demo-readiness/<run_id>",
    )
))
# result.morning_dispatch_plan, result.plan_path, result.human_summary

Architecture

SKILL.md   -> thin shell (this file)
rehearse   -> omnimarket/src/omnimarket/nodes/node_demo_rehearsal/
detect     -> omnimarket/src/omnimarket/nodes/node_demo_drift_detector/
dispatch   -> omnimarket/src/omnimarket/nodes/node_demo_fix_dispatcher/
handoff    -> omnimarket/src/omnimarket/nodes/node_morning_handoff_generator/
evidence   -> docs/evidence/demo-readiness/<run_id>/