The deliverable is complete and verified.

What was built

• src/SatoshiTest.sol is a two-line contract over OpenZeppelin v5.7.0 ERC20. The constructor mints 21,000,000 × 10^18 SATS to msg.sender once. It has no constructor arguments, no owner, no mint, burn, pause, blacklist, fee, hook, receive or fallback, so the launch factory receives the whole supply and every launch flow moves exact amounts.
• script/Deploy.s.sol reads only the optional EXPECTED_CHAIN_ID (0 disables the check, otherwise it must be 31337 or 11155111 and match the live chain), deploys exactly one contract between the broadcast markers, and exposes deploy() and checkChain() for direct testing.
• test/SatoshiTest.t.sol has 30 tests: metadata and supply, CREATE2 deployment by a contract, transfer and approve success paths and events, six distinct revert paths with exact OZ custom errors, the same 10 mint/admin and 12 freeze/seize selector probes the protected floor test uses, an opcode scan for DELEGATECALL/CALLCODE/SELFDESTRUCT, EIP-170 size, the deploy script's guard, and four fuzz tests on supply conservation and allowance handling.
• lib/ holds the OZ ERC20 subset and forge-std 1.16.2 as plain files with licences; no submodules, no remote imports.
• README.md documents rules, manifest values (decimals 18, totalSupply 21000000000000000000000000, empty constructorArgs), offline build steps, the operator-only deploy command, operational responsibilities and assumptions. REVIEW.md records the checklist pass, how each floor check is mirrored, and what was re-run.

Verification results

Check	Result
forge build --offline (solc 0.8.26)	compiles, no warnings
forge test --offline	30 passed, 0 failed
forge fmt --check	clean
EXPECTED_CHAIN_ID=0 forge script …Deploy --offline	runs, one deployment

The protected floor test itself could not be run here since it imports the factory repository's helper libraries, which are not part of this task's inputs. Its token-level checks are reproduced in the delivered suite. Slither and Mythril were not available and did not run.

launch.json is written and validated. The only change in the tree is the new manifest; the other untracked entries in git status are sandbox-mounted device nodes, not files I created.

What the manifest says

• kind is custom_token, token contract SatoshiTest with name Satoshi Test, symbol SATS, 18 decimals, empty constructorArgs and totalSupply of 21,000,000e18 in minor units, which is exactly the INITIAL_SUPPLY the constructor mints to msg.sender.
• contracts is empty: the accepted tree holds only the token and a standalone operator deploy script, no application contracts.
• pool pairs with native ETH at the zero address, fee 3000, tickSpacing 60, matching prior accepted manifests on this box. initialPrice is the provenance sqrtPriceX96 derived from 5 ETH over the full supply with ETH as currency0.
• economics is the job's object copied verbatim.
• notes (1998 chars) describe the token's fixed supply, lack of privileged functions, exact launch flows, and that the deploy script is not used by the factory.

Checks run

• forge build --offline: compiled cleanly with solc 0.8.26.
• forge test --offline: 30 passed, 0 failed.
• Constructor ABI in out/ confirms zero inputs, matching the empty constructorArgs.
• A scratch schema check (test/scratch/check.py) verified every field, pattern, length bound, the economics object, and that totalSupply equals 21,000,000 × 10^18.

Wrote .imd-findings.json.

• No substantiated defects.
• All three entry points covered.
• 30 local tests passed.
• Protected pool/swap tests remain unexecuted because required dependencies and deployment inputs are absent.

No source or configuration files changed.

Recorded no substantiated defects and coverage of all three entry points in .imd-findings.json.

All 30 local tests passed. Protected seed-and-swap integration remains unverified because required dependencies are absent.

Saved .imd-findings.json with no substantiated findings and coverage for all three entry points.

All 30 tests passed, including four fuzz tests. Protected pool-integration execution remains unverified because required dependencies and launch inputs are absent.

Wrote .imd-findings.json with no substantiated defects and coverage for all three entry points.

All 30 local Foundry tests passed. Full protected launch integration remains unverified because required service dependencies and launch inputs are absent. No source or configuration files changed.

Done. forge build and forge test both pass: 61 tests across three suites, 0 failures. No defects were found, so no .imd-findings.json was written.

What I added (two new files, nothing existing rewritten):

• test/SatoshiTest.invariant.t.sol adds the invariant suite the acceptance criteria required. A handler drives the token from six actors with raw and clamped entry points, an adversarial layer that fires unimplemented selectors and plain ether at the contract, and a donation handler that sends tokens to the token's own address. It keeps an independent model of balances and allowances and asserts after every call that the token did what the model predicts or reverted with exactly the predicted error, including OpenZeppelin's allowance-before-receiver-before-balance ordering. Seven invariants cover constant supply, sum of balances equals supply, model agreement for balances and allowances, tokens at the contract address never leaving, strangers holding nothing, and unchanged runtime code with zero ether. Inline config sets 256 runs at depth 64 with fail-on-revert enabled, and the campaign ran 16,384 calls with zero reverts.
• test/SatoshiTest.edges.t.sol pins failure paths the existing suite skipped: maximum-uint amounts, exact-boundary and max-minus-one allowances, the zero address in every argument position, the owner spending from itself without self-approval, approvals overwriting rather than accumulating, transferFrom emitting a Transfer but no Approval, reverted calls emitting nothing, tokens stuck at the contract address, and independence of a second deployment. Fuzz tests at 1000 runs check round-trip, split-versus-single transfer metamorphism across two deployments, and infinite-allowance behaviour.

Verification beyond the green run. A scratch mutant that over-credited the model by one wei failed five invariants with a shrunk two-call sequence, so the suite is not vacuous. I also diffed the vendored OpenZeppelin files against upstream v5.7.0: ERC20.sol differs only in import order and the renamed IERC6093 path, with identical logic, and the other files match exactly.

Not a finding, worth knowing. The invariant suite takes roughly 40 to 120 seconds depending on machine load, which is the cost of 256 runs at depth 64 with model checking in every handler.

No reproducible defects found. Wrote .imd-findings.json with empty findings and all three entry points marked holds.

Offline Foundry validation: 61 passed, 0 failed, including invariant testing. Protected pool integration remains unverified because required dependencies and deployment inputs are absent. Implementation files are unchanged.