ASI Takeoff operator runbook

Start with the local studio and acceptance exercise. The paths below retain the original contract and reporting capabilities while making their outputs and execution boundaries explicit.

1. Prepare the pinned toolchain

Follow the root setup. Use locked dependencies, not npm install or an unpinned foundryup upgrade. Node-only planning and studio commands do not require dependency installation. Full contract runs do.

Default local fixtures need no personal wallet. Do not overwrite an existing .env or copy production keys into the demo. env.example documents optional variables; select only the values needed for this rehearsal. The local launcher rejects non-loopback RPCs and conflicting chain settings and uses chain ID 31337.

Prepare the computer-work surface

Choose the surface before admitting a task. For OpenClaw's native Codex Computer Use, verify the Codex-owned plugin/MCP tools, the target desktop session and OS permissions. An installed plugin alone does not prove the desktop responds: the documented computerUse.strictReadiness option adds a live startup probe. Configure it through the commissioned OpenClaw/Codex environment; this demo does not install or repair that environment.

For ChatGPT Work Computer Use, enable the plugin and review app access in the desktop app. macOS and Windows have different desktop behavior; Windows uses the foreground session. Use approved structured tools for repeatable operations and Computer Use when visual interaction is required. Export the two candidate files and run review.cjs as shown in the README.

For an admitted OpenClaw run, retain the exact adapter receipt and protected dispatch journal. Use review.cjs --receipt with the job and deployment from your admission record. A rejected receipt does not authorize automatic redispatch. Reconcile the original attempt before admitting a correction. The checker validates task/content consistency; it cannot authenticate unsigned provider claims or approve payment.

If CI reports HH502 during compiler download, inspect the active network policy and the pinned compiler source before retrying. This workflow allows binaries.soliditylang.org:443 for Hardhat's compiler lists/binaries while retaining block mode and agent health checks. Do not substitute a compiler version or disable the firewall to obtain a green run.

2. Run the retained three-job local chain

npm run demo:asi-takeoff:local

The launcher compiles incrementally with one compiler job, starts its own loopback Anvil/Hardhat node, deploys v2 defaults, and runs demo/aurora/aurora.demo.ts with demo/asi-takeoff/config/mission@v2.json. It checks final job states, independently reconciles receipt accounting, renders the report and stops its node. The jobs cover agriculture, infrastructure and healthcare with configured results and mock-token transactions. No model provider or real-world project is executed.

If another service occupies port 8545:

DEMO_PORT=18545 npm run demo:asi-takeoff:local

Existing nodes are left running. Ctrl+C stops only the node owned by the launcher. Remove conflicting exported RPC/chain variables rather than directing the rehearsal to another chain. Inspect the printed node log after startup failures. An interrupted chain is disposable; retain its receipts as incomplete evidence and run a new namespace instead of attempting to resume against stale addresses.

Inspect and bundle the actual local receipts

Under reports/localhost/asi-takeoff/:

The report can be rebuilt after the node stops:

npm run demo:asi-takeoff:report
npm run demo:asi-takeoff:kit -- --report-root reports/localhost/asi-takeoff --local-receipts reports/localhost/asi-takeoff/receipts --summary-md reports/localhost/asi-takeoff/asi-takeoff-report.md --network localhost

Local-receipts mode defaults to the three-job config/mission@v2.json, not the separate rail plan. For a custom mission, supply its exact --plan file. The kit hashes deployment, mission, stake and governance files and optional summary bytes. Its directory listing does not recursively hash every receipt, and the kit does not independently query chain state or certify settlement. Preserve the full receipt tree and the commissioning-check output.

3. Run the governance report pipeline

npm run demo:asi-takeoff

This distinct pipeline regenerates constants, compiles, runs testnetDryRun.ts, creates thermodynamic and owner-control reports, checks owner wiring and packages a governance kit. Report steps use separately bootstrapped Hardhat fixtures. Their addresses should not be interpreted as one persistent deployment shared with the local launcher. The national rail plan supplies summary context; its five planned jobs are not dispatched.

Outputs under reports/asi-takeoff/ are dry-run.json, thermodynamics.json, mission-control.md, summary.md, summary.json, mission-bundle/, logs/, governance-kit.{json,md} and run-status.json. Bootstrap steps can also write local .hardhat.json configuration overlays; inspect them before reusing local owner commands. Never copy these addresses to a live deployment.

For planetary context, use the pipeline's actual override and a separate output directory:

ASI_TAKEOFF_PLAN_PATH=demo/asi-takeoff/project-plan.planetary.json ASI_TAKEOFF_REPORT_ROOT=reports/asi-takeoff-planetary npm run demo:asi-takeoff

The output filenames stay the same within that directory. This override does not alter demo:asi-takeoff:local. The plan's historical reporting.artifacts paths are scenario metadata, not routing instructions honored by the pipeline.

Offline fixture mode

AGIJOBS_FLAGSHIP_SKIP_ONCHAIN=true ASI_TAKEOFF_REPORT_ROOT=reports/asi-takeoff-offline npm run demo:asi-takeoff

This mode writes explicitly labeled fixtures. The dry-run status is simulated; its scenario is not-run. No ownership checks, chain transactions or provider work occur. A generated kit remains offline evidence and has productionApproved: false. Use a different report directory for each run. A missing/running run-status.json means the run did not complete; old files may coexist after a failure and must not be taken as fresh results.

4. Owner control and validator oversight

The local launcher stops its node when it exits. Its saved addresses cannot be used for RPC commands against an unrelated new node. Advanced owner previews require a separately managed local deployment with actual configuration overlays and a running RPC.

HARDHAT_NETWORK=localhost npm run owner:verify-control
HARDHAT_NETWORK=localhost npm run pause:test
HARDHAT_NETWORK=localhost npx ts-node --transpile-only --compiler-options '{"module":"commonjs"}' scripts/v2/updateThermodynamics.ts

pause:test inspects wiring and uses static calls; it does not broadcast a pause. Thermodynamics defaults to a preview. The direct invocation uses --transpile-only, matching the repository's Hardhat runtime setup; ordinary ts-node type-checking fails on the existing Hardhat/ethers script types before reaching the preview. Adding --execute broadcasts transactions and requires separately recorded authorization, the correct signer and reviewed parameters. Hardhat's run command does not forward arbitrary --execute script arguments. The kit uses the direct invocation.

The owner command center and dashboard remain available (npm run owner:command-center -- --network localhost, npm run owner:dashboard -- --network localhost) after their actual deployed module configuration is prepared. A saved AURORA_DEPLOY_OUTPUT file alone does not re-create a running chain or automatically populate every owner script's configuration.

The mission receipts include validator selections and commit/reveal evidence. Reproduce a full lifecycle on a fresh local chain using the launcher. Do not replay commit/reveal commands with arbitrary salts or old job IDs from a stopped chain. Commission durable reveal-secret handling and independent evaluation for live computer-work validation; the generic structural evaluator intentionally abstains.

5. Optional Python and Make launchers

The Python wrapper is optional. In a virtual environment:

python -m pip install -r demo/asi-takeoff/requirements-cli.txt
python demo/asi-takeoff/run_demo.py check
python demo/asi-takeoff/run_demo.py run --help

check verifies basic prerequisites/assets, not production readiness or a complete compiler install. run accepts only --network localhost. Relative input/output paths are resolved before the shell changes directories. Namespaces cannot contain path separators. The existing Make commands work from either location:

make -C demo/asi-takeoff local
make -f demo/asi-takeoff/Makefile report

6. Extend, recover and commission

• Change config/mission@v2.json and its specs for local contract jobs. Rewards/stakes retain their original units; validate against the actual token and contract limits.
• Change the project plans for planning/context. Validate both budgets and dependencies with demo:asi-takeoff:plan; clarify deadline semantics before treating them as schedules.
• Retune config/asi-takeoff.thermostat@v2.json only as a reviewed local fixture. It does not prove economic outcomes.
• Admit new computer-work tasks using the protected worker procedure. Do not infer acceptance from provider completion or tool access. Retain original receipts/journals after interruption; reconcile actual effects before another dispatch.
• Preserve reports when comparing runs. Use a new AURORA_REPORT_SCOPE for local jobs or ASI_TAKEOFF_REPORT_ROOT for the pipeline. Archive completed evidence before any intentional cleanup.

Production commissioning requires task-specific evaluation, actual access/identity/authority verification, secure worker configuration, interruption tests, durable journals, audited deployment changes, contract receipts and finality. This repository rehearsal does not itself provide those live approvals.