email-writing
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides an 'Audit Email Performance' feature that instructs the agent to process content pasted by the user. This creates a surface where malicious instructions embedded in a pasted email could potentially influence agent behavior.
- Ingestion points: SKILL.md under the 'Audit Email Performance' section.
- Boundary markers: Absent; usage examples do not include delimiters or specific instructions to ignore embedded directives in the pasted content.
- Capability inventory: No file system writes, network operations, or subprocess executions are defined within this skill.
- Sanitization: None provided; the skill does not instruct the agent to sanitize or validate the pasted email content before processing.
- [METADATA_POISONING]: The 'Skill Boundaries' section contains descriptions unrelated to the skill's primary purpose. It references 'audio production workflows' and 'audio engineering expertise' instead of email marketing, which appears to be a copy-paste error from a template. This inconsistency is misleading regarding the skill's actual scope.
Audit Metadata