The Agent Skills Directory

[SAFE]: The skill's instructions are purely educational and focused on best practices for n8n workflow development. There are no signs of malicious intent, prompt injection, or data exfiltration.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill explicitly advises against using $env for secrets due to security hardening (N8N_BLOCK_ENV_ACCESS_IN_NODE). It correctly identifies that sensitive operations should be delegated to n8n's credential-aware nodes rather than being handled directly in code.
[INDIRECT_PROMPT_INJECTION]: The skill identifies attack surfaces where untrusted data enters the agent context (e.g., Webhook bodies) and provides robust remediation guidance.
Ingestion points: SKILL.md and DATA_ACCESS.md describe ingesting data via $input.all() and $json.body from external triggers.
Boundary markers: The documentation warns users about the structure of external payloads and the risk of undefined properties.
Capability inventory: The skill maps out available capabilities including network requests via this.helpers.httpRequest and data persistence via $getWorkflowStaticData.
Sanitization: ERROR_PATTERNS.md provides specific code patterns for input validation, optional chaining, and guard clauses to safely handle untrusted external data.
[DYNAMIC_EXECUTION]: The skill accurately describes the n8n JsTaskRunnerSandbox restrictions, informing users that certain dangerous functions (like authenticated HTTP requests) are deny-listed by the platform for security reasons. It guides users toward safer, platform-native alternatives.
[OBFUSCATION]: Base64 encoding is mentioned in COMMON_PATTERNS.md, but only as a standard utility for data transformation (e.g., using the Buffer class). No hidden or malicious obfuscated content was detected.

n8n-code-javascript