I do think this is a really good general idea - and certainly fits into "this could be a BIP". Current covenant-less "deadman's switch" style systems are of course limited, but useful, at least to the enthusiast that can hand-craft them. Avoiding the "need to refresh every interval" is clearly desirable, though the extra complexity *may* end up being more trouble than it's worth (clearly, you don't think so!).

Btw, do you think you should address directly the philosophy found in Liana, and what setups they have decided to expose? (i.e. what's the delta, here).

"The testmempoolaccept RPC can receive a list of transactions in which the later transactions may depend on earlier transactions - however in our case the Recovery Transaction has an nSequence relative-locktime, and therefore calling testmempoolaccept 'alert-tx' 'recovery-tx' will fail, claiming that the Alert Transaction UTXO is not confirmed (and the required time window has not passed). "

This sounds like a pretty chunky and important problem (which surprised me, as in, it never occurred to me). Speaking for myself, I would not rest easy with a recovery plan that I couldn't unambiguously check as follows: "I really, really want to know that this recovery works, but much more than that, I *must* know that it doesn't work (leak my funds) before the intended deadline".

With a single nlocktime you can do that (testmempoolaccept -> "non-final").

Is there a case for requesting a feature in Core that one could do a 'testmempoolaccept' with a conditional setting of the time/block height as argument? Would that solve this issue? I'm guessing not, as (clue is in the name!) the tool was mostly designed around "how does this tx interact with the currently seen mempool" more than "forget conflicts, is this even valid/could it ever be valid in and of itself?" (yeah perhaps it's just an incoherent question..?).

"alert_inputs (mandatory): An array of up to 10,000 inputs spent by the Alert Transaction. Each input is a string in the format "txid:vout" where txid is a 64-character lowercase hexadecimal string and vout is a decimal number of up to 6 digits.
alert_tx (mandatory): The raw Alert Transaction in uppercase hexadecimal format. A string of up to 800,000 characters."

I'm probably being an idiot, but: since you're defining alert_inputs as just txid:vout, what's the point of including it since that data is already in alert_tx, no?

On Sunday, December 28, 2025 at 11:59:43 AM UTC-3 Oren wrote:

Reposting here from BitcoinTalk:

After a short talk with Ava Chow during BTC++ Taiwan, I'm starting this thread to discuss whether my idea is BIP-worthy.

Motivation for Timelock-Recovery plans:
Storing seeds for recovery & inheritance is scary.
Pre-signed transactions to a secondary-wallet/custodian, are safer to handle and backup due to their immutability.
A single pre-signed transaction with a future nLocktime requires "renewal" when the nLocktime deadline is getting close, which could be annoying (i.e. if the seed is split over multiple geographic locations).
Covenants/Vaults are still being debated, and could scare less-technical Bitcoiners.

Solution:
Pre-signing a pair of transactions:
• Alert/Initiate Transaction: A consolidation transaction that keeps most funds on the original wallet (except for a minimal amount that goes to anchor-addresses, for CPFP acceleration)
• Recovery Transaction: A transaction that moves the Bitcoin from the consolidated UTXO to the secondary-wallet(s), with an nSequence relative-locktime that gives the user enough time to move the funds elsewhere (assuming they noticed that the Alert transaction was mined, and still have the seed or signed an alternative transaction in advance).

Similar to a single pre-signed transaction with a future nLocktime, Timelock-Recovery plans will not include new funds that are added to the wallet, and will be revoked even if a tiny amount is spent. This mechanism is intended for wallets that are going to remain untouched for a long time.

An example implementation can be found in the Timelock Recovery plugin that I've implemented for Electrum (merged since Electrum v4.6.0b1). Details and demo videos can be found at: https://timelockrecovery.com.
The plugin creates a UI for signing the two transactions, then saving them either in a PDF file (with detailed manual instructions for less-technological Bitcoiners how to broadcast them), or in a JSON format.

The BIP will be about the JSON format, which includes not only the raw transactions themselves, but also user-information (i.e. name, description, destination-labels, wallet-name, wallet-version), and data about the transactions (i.e. txids, amounts, fees, input-utxos, anchor-addresses, relative-locktime).
A standard JSON format will allow implementing a compatible feature on other wallets, as well as apps/servers for monitoring & initiating timelock-recovery plans - such as the one being developed by RITREK.com (disclosure: I'm one of RITREK's founders).

Let me know what you think!

Oren