a6-plugin-http-logger
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill facilitates the transmission of system and traffic logs to external URIs. Captured data includes client IP addresses, request/response headers, and potentially full request/response bodies if explicitly enabled in the configuration.
- [COMMAND_EXECUTION]: The skill utilizes the
a6command-line utility for gateway management tasks, includinga6 route create,a6 route update, anda6 config sync. These commands are used to apply the logger configurations to the APISIX instance. - [CREDENTIALS_UNSAFE]: Examples within the documentation demonstrate the use of the
auth_headerfield with truncated hardcoded Bearer tokens (e.g.,eyJhbGciOiJIUzI1NiIs...). These serve as instructional placeholders for configuring endpoint authentication. - [PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by capturing and exporting untrusted user data (headers and bodies) to external logging backends.
- Ingestion points: Request and response bodies, headers, and URI parameters are captured via NGINX variables (SKILL.md).
- Boundary markers: None are defined in the provided configuration examples to delimit untrusted content from log metadata.
- Capability inventory: The skill uses the
a6CLI to synchronize configurations and create routes (SKILL.md). - Sanitization: No evidence of data sanitization or escaping of the captured log content is provided in the configuration examples.
Audit Metadata