The four requester-held addresses that blocked the first HiveDesk job are now baked into the project as its mainnet launch inputs, with tests and documentation. Build, test and format checks all pass.

What changed

• src/DeploymentConfig.sol gains four constants: the strategy wallet and the three oracle signers, exactly as the brief gives them. The contracts still take them as immutable constructor arguments, which is what the brief's no-owner rule requires.
• script/Deploy.s.sol now exposes a pure function returning the complete mainnet argument set. The run entrypoint forwards it and refuses any chain other than mainnet. It no longer reads environment variables.
• test/LaunchInputs.t.sol is new, with 19 tests. They pin the literal brief values so a drift in the constants fails loudly, check the four are non-zero and pairwise distinct, deploy with the real argument set and verify the wiring, and then prove power goes to those addresses and no one else. Success cases cover the strategy wallet allocating and settling at a profit. Failure cases cover oracle signers and outsiders trying to move vault funds, a strategy logged by a signer being refused, the cap, plain ETH from the strategy wallet, a counted quorum of non-panel signatures, duplicate or zero panel slots, and the script on a non-mainnet chain.
• README.md documents the addresses, what was verified about them, the operational duties of each key, and the key-loss consequences.

Verification results

Check	Result
forge build with solc 0.8.26	clean
forge test with the verifier's fuzz seed	94 passed, 0 failed, 1 skipped (offline fork suite)
forge fmt --check	clean

What I checked about the addresses on mainnet over a public RPC. All four checksums are valid, and all four have no code and nonce 0, so they are fresh externally owned accounts. That is a requirement for the panel, because the oracle verifies signatures with ecrecover, so a smart wallet could never attest. The README states this, and notes that only the requester can vouch for who controls the keys.

Open items for the launch step, not for this job. No launch.json was written, as instructed. The oracle and strategy keys cannot be rotated after deployment, so the README now spells out that losing two panel keys means a redeployment of the oracle alone. The adversarial review before release still stands as recommended in the first job.