The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The code in SKILL.md contains a hardcoded development fallback secret: secret: authSecret || "dev-secret-for-development-only-change-in-production". While intended for local development, hardcoded secrets are a security risk if they are accidentally used in production or checked into version control.
[CREDENTIALS_UNSAFE]: The suggested production database configuration in SKILL.md disables SSL certificate validation: ssl: process.env.NODE_ENV === "production" ? { rejectUnauthorized: false } : undefined. This allows Man-in-the-Middle (MITM) attacks, which could lead to the exposure of database credentials or session data.
[PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by instructing the agent to ingest and act upon data from external MCP tools.
Ingestion points: Data returned by @better-auth:list_files and @better-auth:search tools as described in SKILL.md.
Boundary markers: Absent. The skill does not provide instructions to treat tool output as untrusted or to use delimiters to separate tool output from instructions.
Capability inventory: None. The skill is primarily a collection of documentation and code templates and does not include scripts with execution capabilities.
Sanitization: Absent. No validation or sanitization logic is suggested for processing content from external sources.

better-auth-skill