Treat recovery as an operating question
Imagine you cannot use your everyday signing device tomorrow. What information would let you identify the right wallet, find its official recovery procedure and understand who else can authorize an action? This exercise answers those questions on paper. It is deliberately separate from an actual recovery attempt. Do not wipe a device, reveal a phrase, change a threshold or transfer funds just to complete this worksheet.
The goal is to find a missing dependency while there is time to fix the plan. A sentence such as “I have a backup” is incomplete if you cannot identify which wallet it restores, whether an additional passphrase is required, or whether imported accounts need different material. MetaMask’s documentation distinguishes a wallet recovery phrase, individual private keys and the local app password. Trezor’s documentation explains that backup formats differ. A single generic instruction is therefore not enough for every setup.
Write an inventory that excludes the secrets
Begin with a local wallet label and the account type. Is it a recovery-phrase wallet, a hardware-backed signer, an imported account or a smart account controlled by several signers? Add the software or device family, the networks in use and a trusted route to the official instructions. These fields are not credentials, but the inventory may still be sensitive. It need not be published or emailed widely.
| Field | Useful entry | Keep out of this document |
|---|---|---|
| Wallet identity | Local label and account type | Recovery words and private keys. |
| Backup dependency | Required backup format and whether another factor is needed | The actual phrase or passphrase. |
| Signer role | Who is responsible for a role and how to reach them | A shared copy of every signer’s secret. |
| Official procedure | Documentation title and trusted location | Recovery links received through unsolicited messages. |
| Evidence | Date the plan was reviewed and unresolved questions | A claim that a funded recovery was tested when it was not. |
Distinguish access to information from authority to spend. A person may know where the operating guide is kept without being a signer. A signer may be able to approve routine payments but be unavailable during an incident. Document the role, not merely the person’s name, so changes in the team do not silently invalidate the plan.
Worked example: two of three signers
In ordinary operation, operations and finance review a payment. The reserve signer is available for continuity but is not asked to approve routine activity. The team documents that the three roles use independently recoverable credentials. Three app installations backed by the same recovery secret would not represent three independent recovery paths. The inventory must describe the underlying dependency.
On Monday, the operations device becomes unavailable. Finance and reserve remain able to sign. Two available signers still meet a threshold of two. The team can consider a carefully reviewed replacement of the unavailable signer through the normal account process. Safe’s setup guidance explains this relationship between a threshold below the total signer count and continued operation after one signer loses access.
On Tuesday, consider a different incident: both operations and finance are unavailable. Only reserve remains. One signature does not meet the two-signature threshold. The existence of a third role never promised tolerance for losing any two roles. Writing this case down prevents an emergency from turning into experimentation with a funded account.
| Available roles | Count | Ordinary threshold met? |
|---|---|---|
| Operations + finance + reserve | 3 | Yes. Two are required. |
| Finance + reserve | 2 | Yes. |
| Operations + reserve | 2 | Yes. |
| Operations + finance | 2 | Yes. |
| Reserve only | 1 | No. |
| None | 0 | No. |
The matrix describes authorization capacity only. It does not tell you whether the remaining signers can safely identify a replacement, whether they can access the correct network, or whether their devices share a common point of failure. Add those operational questions below the arithmetic instead of presenting a green “yes” as a complete recovery guarantee.
Test independence, not just quantity
Ask what could make several roles unavailable at once. Examples for the exercise include all devices stored at one office, all instructions locked behind the same unavailable account, or all signers depending on the same person to explain a transaction. These are hypothetical prompts to examine your arrangement, not claims about any provider’s security.
For each prompt, write a factual answer and an unresolved question. “Reserve device is elsewhere” may answer the storage question but not whether its backup is usable or its operator is reachable. “Two signers understand the process” may still leave uncertainty about what happens when one leaves the organization. A useful drill surfaces those gaps without collecting the secrets needed to exploit them.
Separate an unavailable signer from a suspected compromised signer. An unavailable device creates an access problem. A compromised credential may create an authorization problem even while you can still use it. The decision, urgency and verification needed are different. This worksheet does not provide a one-click incident response. Use the current official product guidance and competent help for an actual event.
Recovery features are another authority path
Some smart-account setups offer a separate recovery route. Safe’s RecoveryHub documentation describes recoverers that can propose signer and threshold changes, subject to a configured delay. It also makes clear that this feature must be configured before the loss. A recoverer is therefore not merely an emergency contact. The role has significant authority that belongs in the inventory.
For a configured recovery route, record the role that can initiate it, the delay, any execution window, how pending changes will be noticed and who can challenge an unauthorized proposal. Do not copy a default delay from an article into a plan as if it were the account’s actual setting. The reviewed documentation may describe one implementation while the live account uses another configuration.
Now run the paper scenario again. If the normal threshold cannot be met, is there a previously configured and independently accessible route? If the answer is unknown, the drill remains incomplete. If the answer is no, say so. A written promise to contact support does not create a recovery mechanism that the account never had.
How to conduct the tabletop exercise
Choose a facilitator who reads the scenario and a recorder who captures the answers. Give each participant only the operating information their role should normally have. Do not circulate backup phrases or ask people to demonstrate a secret on a shared screen. The exercise should reveal whether the written plan and roles are understandable without improvising credentials.
Use four prompts: the everyday device is unavailable. One signer cannot be reached. The location holding a backup is inaccessible. And a claimed recovery request arrives unexpectedly. For each prompt, ask who may act, what evidence they need, what action remains impossible and when the team would stop. Record “unknown” whenever the group cannot establish an answer.
Finish with an action list that names a responsible role and a review date. An action may be to clarify the backup format, update a trusted documentation link, replace an outdated contact route or review the actual account configuration. Those are improvements to the plan. They are not proof that recovery has been successfully executed.
A completion record you can reuse
Download the blank recovery drill sheet (TXT)
The worksheet contains no account addresses, secrets or real participant details. Save a copy locally and record the scenario, available roles, required threshold, independent dependencies, unresolved questions and follow-up owner. Keep a version date so a later review can explain what changed.
A useful completion statement is precise: “We reviewed the two-of-three availability cases on paper and identified two unresolved dependencies.” It is more informative than “recovery is safe.” If you later perform a separate recovery test, document its exact scope and result without exposing the recovery material. Read the recovery overview for product differences, and the Safe assessment for shared-approval context.
Source documents
Official documentation reviewed November 11, 2026. Product terms and availability can change.