ai-agent-builder

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs agents to search and read files (including .env) and even shows an example revealing an API key (CLAUDE_API_KEY=sk-...), and its tools return file/command output as raw strings to the LLM, which encourages exposing secrets verbatim.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill exposes unrestricted file read/write and arbitrary shell execution tools (read_file, write_file, edit_file, run_command), demonstrates searching for secrets (ReAct .env example), and includes devops/deployment patterns without safeguards, enabling modification of system files, running privileged commands, or creating accounts.