A cold-start test gives a fresh builder one bounded task, restricts help to the recorded handoff, and treats every unanswerable question or failed command as evidence about the transfer.
Use a genuinely fresh starting point
The person or agent running the test should not have participated in writing the handoff. Start in a clean clone with the supported runtime and only the access listed in the environment contract. Record the starting commit, operating system or container, commands used, and any choice that required outside explanation. This prevents local memory from quietly filling gaps in the repository.
The first task should be small enough to finish within a few hours but broad enough to cross the normal path from issue to check. A focused defect, an isolated feature, or a documentation change with an executable example works better than an open-ended architecture task. The task specification should state the expected behavior and the checks that define completion.
Record questions before answering them
When the builder asks which command to run, which service owns a field, or whether a failing test is expected, record the question and the place where the answer should have been found. Answering immediately may be necessary to continue the test, but the handoff still failed at that point. The repair belongs in the repository, linked from the surface where a future builder will look.
Google's SRE workbook describes engineers reading handoffs, consulting service documentation, and practicing operational scenarios before assuming responsibility. The same mechanism applies to a repository transfer: written context and rehearsal work together, and the rehearsal exposes gaps that a document review will not.
Require the final state to reproduce
After the task passes, repeat the setup and validation from another clean environment or runner. The receipt should include the final commit, changed documentation, commands, test results, unresolved questions, and access that was deliberately unavailable. A handoff passes when the evidence can be followed again, not when one guided session reaches a working screen.
Cold Start Ledger supervises and records this test through Reality Contact, LLC. The customer chooses the task, confirms the business decisions, and decides whether the resulting pack is ready for the next owner.
Where the service stops
Reality Contact, LLC prepares and verifies the handoff, but does not maintain or operate production, retain credentials, decide undocumented intent, or present guesses as historical decisions. The buyer reviews the pack, confirms the recorded decisions and access boundary, gives it to a fresh agent or developer, and authorizes that person to complete one bounded task from the verified state. This is technical documentation and verification support; it does not replace legal, security, employment, intellectual-property, or production-operations review. We do not promise that a repository is defect-free, that every historical decision can be recovered, or that a new builder can work without questions.
Sources: Google SRE workbook chapter on on-call handoffs and playbooks; Git documentation for cloning a repository.