Create the DataPipe experiment and a draft Prolific study — critical path

[htsukamoto5]

Phase 1 — unblocked, minutes of work, and currently the binding constraint on two workstreams.

Not a coding task. Filed because it is the critical path and would otherwise exist only as a note: several issues are now blocked on credentials and identifiers that do not exist yet, rather than on any remaining design question.

What is blocked on what

Blocked	Needs
#3 — data egress cannot be built or tested	A DataPipe experiment ID and a linked OSF project
#14 — cannot be tested end to end	The four Prolific completion codes
#7 — codes are placeholders in CONFIG	Same
#5, #6 — payment paths cannot be verified	Same

Work

• Create an OSF project (or component) for the study, and link it to a DataPipe account.
• Create the DataPipe experiment, obtain the experiment ID, and paste it into CONFIG.DATAPIPE_EXPERIMENT_ID.
• Leave limitSessions off, or size maxSessions to the expected save count rather than the participant count — see READINESS.md §3. Sizing it to participants caps the run about a seventh of the way through and drops every subsequent save silently.
• Confirm a test save round-trips to OSF, and that a second save with a colliding filename fails as expected (OSF 409) — this is the one failure path that is not retried, so it is worth seeing once deliberately.
• Create a draft Prolific study (does not need publishing) to obtain the four completion codes, and paste them into CONFIG.COMPLETION_CODES.
• Verify a completion-code redirect registers as a real submission.

Note on ordering

The draft Prolific study also forces early decisions on several publish-locked settings — reward, advertised duration, screen-out reward, slot limit, prescreeners. Those are group B in DECISIONS.md and cannot be changed once published, so it is worth walking that list while creating the draft rather than after.

The OSF project created here is also where the pre-registration should be registered (READINESS.md §7), so linking them now avoids doing it twice.