n8n-expression-syntax
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill documents techniques for interpolating untrusted external data directly into workflow operations, which constitutes a surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the agent context through $json.body (webhook inputs) and references to previous node outputs via $node["Node Name"].json as described in SKILL.md and EXAMPLES.md.
- Boundary markers: The documentation and examples (e.g., Examples 1, 2, and 3 in EXAMPLES.md) lack delimiters or explicit instructions to treat embedded content as untrusted data.
- Capability inventory: The skill teaches how to use these variables in high-impact capabilities including SQL query construction (Postgres node in EXAMPLES.md), HTTP request URL generation, and messaging (Slack node).
- Sanitization: The provided examples do not include sanitization, escaping, or validation steps for the interpolated data before it is used in queries or external requests.
Audit Metadata