01
Identify the input set and preserve its boundary
List the source folder and the kinds of files the job is intended to accept. State whether subfolders are included and how duplicate base names will be distinguished. Treat the source collection as read-only for this job and specify a separate destination for generated previews. Record any formats or special cases that need an explicit decision before processing. The brief should identify who can resolve those questions. A clear input inventory makes it easier to notice an omitted file without treating every item in an unrelated directory as part of the task.
02
Write output rules that another person can interpret
Define the preview dimensions, whether proportions are preserved and the requested output format. Add a naming example, such as turning sample-photo into sample-photo-preview, with a rule for avoiding collisions. Specify what should happen if the destination already contains that name; silently overwriting an earlier file should not be an accidental default. Keep decisions about metadata and colour handling visible rather than assuming every conversion preserves them. A reusable specification row can contain source identifier, output name, dimensions, format and handling notes, leaving unresolved choices marked for review.
03
Separate generated outputs from unresolved exceptions
After an implementation runs, its actual report should distinguish created previews, skipped inputs and failed conversions. That report is future evidence, not something supplied by this planning example. Define the checks the reviewer will perform, including matching the output list to the intended inputs and opening selected previews to inspect their appearance. Record any exceptions and the decision about them before repeating the job. Retain the specification version alongside the output list so a later reviewer can tell which rules produced the copies and whether a changed brief requires a new run.