best-practices-converse
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill defines policies for conversational tone, emotional steering, and paralinguistic cues in voice-first agent interactions.\n- [SAFE]: The instructions mandate the sanitization of user-supplied control characters (brackets/XML) to prevent prompt injection or unintended behavior in the Text-to-Speech (TTS) engine.\n- [SAFE]: Potential indirect prompt injection risks are addressed by requiring input turns to be resolved and intent-classified before memory recall, with a specific rule against passing user-supplied tags to rendering systems.\n- [COMMAND_EXECUTION]: The evaluation fixture (fixtures/agentic_eval.json) includes a command to run a local validation script (../../best-practices-skills/scripts/validate_skill.py). This is a static analysis tool used for skill contract verification within the development environment.
Audit Metadata