# Archive Signal — Sample Capture Brief

Illustrative scoping document. Not an available dataset, quotation or collection commitment.

## 1. Capability gap

Distinguish several causes of an intermittent pump fault and choose the next diagnostic test. Preserve the expert's reasoning, including hypotheses that were rejected.

## 2. Intended use and embodiment

Example use: evaluation of diagnostic planning from human-work observations.

Target embodiment, robot action requirements and any transfer from human demonstrations must be specified separately. Human observation video does not itself supply robot joint states or action trajectories.

## 3. Episode boundaries

Begin with the symptom, known equipment state and available service history. Include the expert's observations, chosen tests, decision points, intervention and outcome verification. Record failed interventions and recovery where they occur.

## 4. Candidate signals

- POV and fixed third-person video with useful views of the task.
- Audio and expert explanation at agreed, safe points in the workflow.
- Available diagnostic logs, pressure or temperature readings.
- Redacted work-order context and final inspection or test results.

Signal access, sample rates, camera placement, synchronization tolerance, calibration and missing-data limits must be agreed after a feasibility check. Do not assume that every modality is available.

## 5. Coverage and volume

Describe desired equipment variants, operators, sites, wear, environmental conditions, failure types and unsuccessful attempts. Identify held-out evaluation requirements. Define pilot episode counts and larger-program volume only after event frequency and access are understood.

## 6. Acceptance and rejection

Accept when the relevant sequence is observable, required signals meet the agreed quality bar, and the outcome can be linked to supporting evidence.

Reject or flag when key views are missing, alignment is outside the agreed tolerance, records are duplicated, permissions are incomplete or the result cannot be verified. Preserve uncertainty rather than inventing a success label.

## 7. Outcome verification

Agree the evidence that counts: before/after measurements, diagnostic results, an inspection, return-to-service status or another task-appropriate result. Identify who reviews it and how unresolved cases are labelled.

## 8. Rights and capture boundaries

Identify data controllers, participants and any third-party rights. Agree informed consent, excluded information and locations, permitted uses, exclusivity if requested, confidentiality and any licensing terms before capture.

No hidden recording. No incidental customer capture. No upload of confidential records during the initial inquiry.

## 9. Delivery and security

Specify an episode manifest, modality files, timestamps, labels, provenance, quality notes and rights documentation. Agree schema, file formats, storage region, encryption, access, retention and redaction requirements. Robot-specific delivery formats apply only where the underlying data supports them.

## 10. Commercial scoping

State the desired pilot window, budget range and procurement constraints. Timing and price remain subject to a mutually agreed scope, access and technical feasibility.

## Start your own brief

https://archivesignal.ca/request

Or send a non-sensitive overview to contact@archivesignal.ca. An inquiry is not a data licence or a purchase commitment.