voice-checker
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs its intended function of text analysis using a limited set of read-only tools (Read, Glob, Grep). It does not request network access, write permissions, or administrative privileges. The instructions emphasize an advisory role where findings are suggestions and no automated changes or pipeline blocks occur.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external file content provided via $ARGUMENTS, creating a data ingestion surface where malicious text in the target files could attempt to influence the agent's analysis.
- Ingestion points: Target track, album, or prose files specified at runtime via the $ARGUMENTS variable in SKILL.md.
- Boundary markers: None; the skill reads the files directly without surrounding delimiters or instructions to ignore embedded prompts.
- Capability inventory: The skill is limited to file system read and text search tools (Read, Glob, Grep); it lacks write permissions, shell execution, or network access.
- Sanitization: No explicit validation or sanitization of the input text is performed.
- Assessment: The injection surface is inherent to the tool's core function. Given the absence of active capabilities like network exfiltration or system modification, the risk is minimal and restricted to the integrity of the generated advisory report.
Audit Metadata