# ASTRA prospective case-study protocol

Status: blank template. Not a completed study or approved protocol.

## Accountability and registration
- Question and intended user:
- Responsible researcher / expertise:
- Independent reviewer / expertise / relationship / conflicts:
- Registration date and external versioned artifact URL:
- Consent, data permissions and ethics requirements:

## Four gates — evidence required, no automatic pass
- Useful: audience hypothesis, observed recurring task, current tools, friction, switching reason, de-identified observation artifact:
- Dependable: save/reopen, cross-device access, permissions, original links, review/export and recovery test results; unresolved failures:
- Better: prospective comparison below; quality and error criteria must hold before interpreting speed as improvement:
- Viable: repeat use, repeat-use barriers, purchasing authority, stated willingness versus actual purchase, delivery/support costs and evidence gaps:
- For each gate: pending / blocked / ready for human review; responsible person; next test; decision rationale:
- A 3–5 person usability study identifies friction; it cannot establish population-wide effectiveness.
- Ninety days is a planning window. Advance when evidence supports it, not on a calendar deadline.

## Evidence and decision
- Search databases, exact queries, dates and coverage limits:
- Inclusion/exclusion rules and reasons:
- Reference-set construction and independent checks:
- Source versions, exact passages and locations:
- Competing evidence and unresolved objections:
- Initial decision and conditions that would change it:
- Proposed test, raw-data artifact and reproducibility procedure:

## Optional comparative evaluation
- Named comparator, version and equal training:
- Participants and matched tasks; sample-size justification:
- Randomization/counterbalanced order and external blinding:
- Primary outcome and timing method: separate setup, search, reading, checking, correction, failed searches and reviewer person-minutes for each tool; no double counting; disclose waiting:
- Quality rubric, acceptable quality difference and error definitions:
- Failed/incomplete tasks, exclusions, dropouts and stopping rules:
- Analysis of paired results, participant clustering and uncertainty:
- Preregistered success thresholds and rationale:

## Optional monitoring evaluation
- Working unattended scheduler and authorized access (not supplied by ASTRA):
- Observation period, expected cadence and missed-run handling:
- Independent relevant-paper/correction reference set:
- Precision, recall, false alerts, coverage and detection-delay analysis:

## Results and release (complete after collection)
- Actual dates, deviations, raw artifacts and negative findings:
- Reviewer findings, disagreements and reproduction outcome:
- Revised decision and remaining uncertainty:
- Separate public-release decision, permissions and contributor consent, correction route and limitations (default private; invited sharing is a separate choice):

Keep the preregistered version. Append amendments; do not replace it after seeing results.