medium-style-blog-writer
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it is designed to process untrusted data from multiple sources. 1. Ingestion points: The agent is instructed to research primary documentation from the web and ingest user-provided notes, screenshots, and URLs (SKILL.md, Workflow Steps 1 and 2). 2. Boundary markers: The instructions do not define specific delimiters for external content, though they emphasize treating user notes as a source of truth. 3. Capability inventory: The skill utilizes web search for research and suggests running or reconstructing command sequences in a sandbox environment (SKILL.md, Step 2). 4. Sanitization: The skill mitigates risks by instructing the agent to prioritize primary documentation, avoid inventing data, and keep tutorial steps conservative.
- [COMMAND_EXECUTION]: The instructions suggest verifying tutorial steps by running sequences in a sandbox or connected tool. This is a legitimate quality assurance task for technical content generation and does not involve unsafe execution methods.
- [SAFE]: No evidence of malicious behavior, obfuscation, data exfiltration, or persistence mechanisms was identified. All external resource references are consistent with the skill's stated purpose and author context.
Audit Metadata