The CLI for building test messages is not implemented, but the functions for building test messages are implemented in lib/preparer.ts. An example of the usage of these functions can be found in the file lib/prepareTestData.ts

It looks solid to me.

One thing that seems a bit quirky: in lite client, the sender needs to know the "signature_index" of each validator. Is there a guarantee that we can deduce the signature_index of any master chain block's signatures, or does the sender need to extract this from the specific key block?

Lite-client.
In `set_new_data` function you calculate target_weight as `muldivc(total_weight, 2, 3)`. You use ceil round.
In `check_signatures` function you use `throw_unless(error::signatures_weight_not_enough, total_weight >= target_weight);`

Example. If total_weight == 12, than target_weight == 8.
Let in `check_signatures` total weight of signatures is 8.
In you solution is good block.

But it's bad https://github.com/ton-blockchain/ton/blob/master/crypto/block/check-proof.cpp#L659

So, you should use `muldiv(total_weight, 2, 3) + 1`

Transaction checker.
It's bad idea to handle bounced messages. Attacker can send internal message to your contract with op == 0xffffffff and clean all pending queries

Transaction checker.
It's really interesting that you give `path` in `ShardAccountBlocks` that is HashmapAugE.
You don't analize HashmapAugE. You just go through cells by refs, using `path` =)
But attacker can take random cell from `ShardAccountBlocks`. And your contract answers that is ok transaction. But it's not a transaction.

Solid implementation of lite-client and tx-checker, but "hints"-based tx-search in proof (insecure).

More info on issues: https://contest.com/docs/TrustlessBridgeChallengeAssessment