nielsen-heuristics-audit
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from external sources, creating a surface for indirect prompt injection attacks.
- Ingestion points: The skill accepts
screenshots_or_links,known_issues(bug reports/complaints), andcompetitive_context(competitor pages) as input (SKILL.md). - Boundary markers: The skill explicitly instructs the agent to use boundary markers, specifically scoping external content within
<untrusted-content>tags, and mandates that audit instructions take precedence (SKILL.md). - Capability inventory: No dangerous capabilities such as file-system writes, subprocess execution, or manual network operations were identified in the skill instructions.
- Sanitization: The skill provides clear instructions to the agent to detect and flag injection patterns (e.g., "ignore previous instructions") and to disregard HTML/Markdown formatting or encoded characters in untrusted inputs (SKILL.md).
- [PROMPT_INJECTION]: A static detector identified phrases such as "ignore previous instructions" and "disregard your task" in the skill metadata. However, contextual analysis confirms these strings are part of a 'Security Notice' section intended to teach the agent how to identify and ignore adversarial content in external data. This is a defensive implementation and does not constitute a malicious attempt to override agent behavior.
Audit Metadata