work-order-writeup
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONINGNO_CODE
Full Analysis
- Indirect Prompt Injection Surface: The skill processes external data from a JSON payload and interpolates it into a markdown template to generate work order reports. If the input data contains malicious instructions, it could theoretically influence the agent's output.
- Ingestion points: Untrusted data enters the context via the
work-order-packet.jsonfile, specifically through fields likeproblem_descriptionandsummary. - Boundary markers: The
work-order-writeup-template.mdfile does not currently utilize explicit delimiters or instructions to ignore embedded commands within the interpolated fields. - Capability inventory: The skill is designed to generate markdown, DOCX, and PDF files based on the provided templates.
- Sanitization: The instructions do not specify explicit sanitization or escaping procedures for the external content before it is rendered into the final document.
- Mitigation: The skill explicitly requires that all generated artifacts be marked as "DRAFT" and mandates a human review by a planner or supervisor before any action is taken.
- Metadata Inconsistency: There is a discrepancy between the author listed in the skill metadata ("Microsoft Manufacturing Skills") and the copyright holder named in the license file ("Sravani Seethi"). This is likely a template artifact but should be reviewed for administrative accuracy.
Audit Metadata