blog-style
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local script 'scripts/style_learn.py' using shell commands. The '' argument is user-supplied, creating a potential surface for command injection if the agent does not sanitize the input. Note that the referenced script file was not provided for security analysis.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from blog posts which could contain hidden instructions or malicious payloads. 1. Ingestion points: External markdown, text, or directory paths provided via the '' argument. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded content are present in the prompt template. 3. Capability inventory: The skill executes local scripts and performs file writes (e.g., to VOICE.md or voice-profile.json). 4. Sanitization: No sanitization or validation of the ingested text content is described. Mitigation: Wrap external content in delimiters and sanitize file paths before execution.
Audit Metadata