asterisk-tester

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill includes an inline curl example with basic-auth credentials (curl -u admin:password) and asks for log excerpts, which encourages embedding plaintext credentials in commands or outputs and could cause the LLM to emit secrets verbatim.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, the runtime workflow takes the outsider’s free-text request (“Simply describe what you want to test”) and uses it to drive Asterisk CLI actions like dialplan show <context> / channel originate Local/<extension>@<context> and logging (tail -f /var/log/asterisk/full), meaning attacker-authored text directly influences what the agent reads/execut es at runtime.