The Agent Skills Directory

[DATA_EXFILTRATION]: The skill implements an indirect prompt injection surface by ingesting and processing untrusted external content.
Ingestion points: External content is loaded in 'Phase 1: IDENTIFY TARGET' (Step 2) to be validated and refined.
Boundary markers: The instructions do not specify the use of delimiters (e.g., XML tags, triple backticks) or 'ignore embedded instructions' warnings when handling the target content.
Capability inventory: The skill has access to high-privilege tools including Bash, Write, Edit, Task, and Skill, which could be exploited if the agent follows instructions embedded within the processed text.
Sanitization: There is no mention of sanitizing, escaping, or validating the input text before the agent 'addresses violations' or 'applies fixes' using its tools.
[COMMAND_EXECUTION]: The skill includes Bash in its allowed-tools list. While the instructions focus on text editing and validation, the availability of a shell environment alongside the processing of user-supplied text increases the potential impact of indirect prompt injection.
[SAFE]: No hardcoded credentials, malicious remote code execution patterns, or persistence mechanisms were detected in the skill logic. The metadata and implementation align with the stated purpose of voice fidelity validation.

voice-validator