ZMIME
Software development

Application security reviews AI prompts.

Review the supplied authorized application code and threat context.

Replace the bracketed fields with your information before using a prompt.

Start the task

Act as a software engineer.

Task — Review the supplied authorized application code and threat context.

Context — [describe the situation and intended audience].
Required information — Repository excerpts, runtime, versions, requirements, constraints and expected behavior.
Source material — [paste the relevant facts, text, data or documentation].
Constraints — [length, format, dates, budget, terminology and requirements].
Deliver — Concrete weaknesses, affected flows and defensive remediations.

Use the supplied facts. Identify missing information before relying on it. Preserve names, figures, units and qualifications. Mark assumptions and use placeholders for facts that need confirmation. Ask focused questions if a missing fact would materially change the result.

Improve a draft

Review the following draft as a software engineer.

Original task — Review the supplied authorized application code and threat context.
Required deliverable — Concrete weaknesses, affected flows and defensive remediations.
Audience and constraints — [paste the requirements].
Source material — [paste the facts or references].
Draft — [paste the draft].
Requested changes — [describe the improvement].

Check factual support, completeness, clarity and format. Return the revised draft and a short list of material changes. Mark unresolved questions and retain verified facts.

Check the result

Evaluate this result against the supplied requirements.

Task — Review the supplied authorized application code and threat context.
Expected deliverable — Concrete weaknesses, affected flows and defensive remediations.
Requirements and evidence — [paste the brief and sources].
Result — [paste the output].

Return a table of requirement, evidence in the result, any issue and suggested correction. Check calculations, citations, consistency and missing inputs where relevant. Finish with the specific items a human should verify before using the result.