startup-problem-inventory
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, prompt injections, or obfuscated code were detected in the skill instructions.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from external sources, such as user-provided backgrounds and information gathered via web browsing tools (forums, reviews, etc.).
- Ingestion points: User-provided context (Founder background, Domains) and external research gathered from communities (Reddit, Slack, Discord, GitHub).
- Boundary markers: The instructions do not define specific delimiters for untrusted data but provide a clear workflow for extracting "raw signals" rather than following instructions.
- Capability inventory: The skill can create a standalone
.htmlfile using filesystem access. - Sanitization: The skill contains a specific security instruction to "HTML-escape user-provided text and do not execute or embed user-provided HTML or script," which effectively mitigates cross-site scripting (XSS) risks in the generated artifacts.
- [COMMAND_EXECUTION]: The skill instructions allow for the creation of a local HTML file if filesystem access is available. This is an expected behavior for generating reports and does not involve execution of arbitrary system commands or unauthorized network access.
Audit Metadata