[New Mission] Operative NextGen: Generate an interview prep document

[aprildunnam]

[New Mission] Operative NextGen: Generate an interview prep document

Summary

Author the mission where learners draft a Python-based skill, generate a structured interview prep document from candidate and role data, and prove deterministic behavior across repeated runs. This replaces the classic document-generation prompt with reviewable code and explicit inputs and outputs.

The mission must validate the Python execution surface and supported document libraries in Copilot Studio before prescribing implementation details. Duplicate checks found no issue proposing this complete Operative NextGen mission.

Proposed mission

Field	Value
Section	Operative NextGen
Folder	docs/operative-v2/06-interview-prep-document/
Title	Mission 06: Generate an Interview Prep Document
Operation codename	OPERATION CODE ASSEMBLY
Difficulty	3
Time	60 minutes

Scenario

Recruiters need a consistent interview pack built from the confirmed candidate, job role, and evaluation criteria. The learner creates a Python skill that transforms explicit structured inputs into a repeatable document with candidate context, tailored questions, and a scoring rubric.

Use a document-assembly analogy: the agent selects approved source material, while Python applies the same blueprint every time.

Mission objectives

• Define a stable interview-prep input and output contract.
• Draft and test a Python document-generation skill.
• Retrieve candidate, role, and criteria data through supported tools.
• Generate and store a usable interview prep document.
• Prove deterministic structure and data fidelity across runs.

Prerequisites to state in the mission

• Completion of Missions 01–05.
• A confirmed Candidate and Job Application in Dataverse.
• An Interview Prep Agent connected to the Hiring Agent.
• Access to the supported Python or code-execution capability.
• Permission to store the generated file in the selected destination.

Proposed lab outline

Define the document contract

Specify required inputs, schema validation, missing-data behavior, output format, section order, file naming, and storage destination. Separate deterministic layout from any optional generated question content.

Draft the Python skill

Create a versioned Python skill or supported code artifact that validates inputs and produces candidate summary, role requirements, evaluation criteria, tailored interview questions, evidence prompts, and scoring rubric. Use only libraries confirmed in the runtime.

Connect the data and output

Retrieve the selected Candidate, Job Application, Job Role, and Evaluation Criteria through Dataverse MCP. Generate the document, store it through a supported file or Dataverse operation, and return an identifier or link.

Prove determinism

Run the same fixed input at least three times. Compare section order, headings, source fields, scoring scale, and file validity. Document which content is deterministic and which content may vary by model.

Acceptance criteria

• Validate the current Python/code execution capability, package availability, file-output support, limits, and naming before authoring steps.
• Provide a complete working code asset, not pseudocode.
• Define and validate structured candidate and role input schemas.
• Keep factual candidate and role fields identical to source records.
• Produce a valid document with required sections in a fixed order.
• Never include protected characteristics or unsupported inferences.
• Store the file in a destination already represented in the solution or explicitly add the required component.
• Run deterministic comparison tests and publish the results and tolerances.
• Include failure tests for missing data, invalid schema, and unsupported output.
• Preserve <mission-meta /> and the Operative Mission 06 analytics tag.
• Follow repository writing and validation standards.

Out of scope

• Free-form document generation without a schema.
• PDF generation unless the validated runtime supports it.
• Email delivery, which belongs in Mission 07.
• Production document retention and records-management policy.

Risks

• The runtime might not include python-docx or permit package installation.
• “Python skill” naming and capabilities may change before implementation.
• Generated question wording can remain nondeterministic even when layout is fixed.
• File size and Dataverse storage limits can affect output.
• Candidate data in generated documents requires careful access control.

References

• Current document-generation mission
• Operative v2 document preview
• Skills in Copilot Studio
• Dataverse file columns