Manual Mode is the VALP path for teams that want the evidence discipline before they install or build a compatible runtime.
Manual Mode is valid for learning, design review, external review workflows, and temporary audit trails. It is not Full Mode: it cannot prove automatic dispatch submission, runtime status waits, pane submit proof, or runtime-backed receipts.
Auto Visible Mode can be layered over Manual Mode only for intake. For example, a project rule may automatically create a task folder, but it cannot choose a Leader or author assignments. After the user selects a Leader and the Leader’s declaration is validated, a human still has to copy work, attest delivery/results, and provide evidence. That task must not claim Full Mode proof.
Use Manual Mode when:
Do not use Manual Mode to claim:
task.md with goal, scope, expected evidence, and approval risks.assignment-declaration.json with role reasons.assignment-validation.json plus routing.json.dispatch.md files visibly.dispatch-receipts.jsonl.final-synthesis.md and run valp audit.Manual Mode should use labels that do not pretend to be runtime proof:
manual_dispatch_written
manual_delivery_attested
manual_result_attested
manual_blocked
Meaning:
| Label | Meaning |
|---|---|
manual_dispatch_written |
A dispatch file was written |
manual_delivery_attested |
A human attests the dispatch was copied or sent |
manual_result_attested |
A human attests the expected evidence was received |
manual_blocked |
Delivery or result evidence could not be produced |
These labels are useful audit records but do not equal Full Mode runtime
receipts such as dispatch_submitted.
See:
examples/minimal-task/
Run:
bin/valp audit examples/minimal-task
The example is intentionally small. It shows how a task can preserve evidence without claiming Full Mode runtime proof.