cowork-use-case-one-pager
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [Processing of Untrusted Input]: The skill ingests data from conversation history, user descriptions, and attached files (ingestion points in SKILL.md) to populate a use-case template. While processing external data is a standard practice, it creates a surface for indirect prompt injection. The skill mitigates this by treating all extracted content as untrusted text and implementing a strict HTML escaping protocol (sanitization in SKILL.md) before the data is written to the final output file (capability in SKILL.md).
- [Dynamic Document Assembly]: The skill performs text substitution on a predefined HTML template (assets/template.html). To prevent the execution of potentially unsafe code when the resulting file is viewed in a browser, the instructions require the agent to verify that no script tags, event handlers, or data/javascript URLs are present in the final generated document.
- [Privacy and Data Handling]: The skill includes specific guidance to prevent the accidental exposure of sensitive information. It encourages the use of anonymized descriptions and mandates that personal or confidential data only be included if explicitly cleared for the target audience.
Audit Metadata