generate-student-artifact

Generate Student Artifact

Use For

• Creating synthetic post-encounter notes from case materials.
• Creating synthetic encounter transcripts from case materials.
• Creating timestamped synthetic observation logs only from a supplied or already generated encounter transcript.
• Modeling learner strengths, omissions, misunderstandings, or performance bands.
• Producing examples for later dry-run grading or rubric evaluation.

Do Not Use For

• Grading the artifact or producing a grade sheet; use grading-dry-run.
• Evaluating rubric quality from dry-run results; use evaluate-dry-run.
• Drafting, importing, reviewing, or transforming rubrics; use post-encounter-note-rubric, rubric-import, osce-rubric-review, or osce-rubric-transform as appropriate.
• Creating raw audio, video, images, or non-text evidence.
• Inferring visual or procedural observations directly from case materials without a transcript source.
• Adding hidden case facts that are inconsistent with the supplied case.

Sibling Sequence

Use generate-student-artifact after rubric drafting, import, review, or transformation when a realistic learner output is needed for testing. Use grading-dry-run to score the generated artifact, evaluate-dry-run to analyze the validated grade sheet for rubric improvements, content-validation for disputed clinical, safety, fairness, or severity concerns, and osce-rubric-transform to apply accepted rubric edits.

Workflow

1. Read the case materials, requested artifact type, learner profile, optional target performance band, optional template, any rubric focus, and any supplied transcript.
2. Load the artifact generation contract before producing the final artifact.
3. Load built-in templates when the user asks for a note, transcript, or observation-log template, does not provide one, or requests template selection.
4. If case facts conflict, preserve the uncertainty in metadata and avoid resolving it by invention.
5. For observation_log, use only a supplied or already generated transcript as the source. If no transcript is available, ask whether to generate a transcript artifact first, or generate that transcript artifact instead of the observation log when the user's request clearly allows it. Do not return both artifacts in one output object.
6. Generate one text artifact that intentionally reflects the requested strengths and weaknesses.
7. Return YAML or JSON with artifact_type, artifact_text, and metadata.
8. Run validate_student_artifact.py on saved JSON or YAML when a file output is requested or when validating a generated artifact.

Quality Rules

• Keep the artifact plausible for the stated learner level and performance band.
• Make omissions visible through absent, vague, incorrect, or incomplete text rather than explanatory labels inside the artifact.
• Put assumptions, intentional strengths, intentional weaknesses, case summary, learner profile, and limitations in metadata only.
• For transcripts, keep speaker turns explicit and use only text evidence.
• For observation logs, write concise timestamped observer-style rows derived from the transcript; do not claim native audio/video inspection.
• For notes, write the artifact as the learner would submit it, not as a rubric or answer key.

References

• Load the artifact generation contract for required output fields and validation rules.
• Load built-in templates for built-in note, transcript, and observation-log templates.
• Load the rubric schema only when rubric focus is supplied and field semantics matter.