tech-design-writer

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides explicit shell command templates using curl for the agent to execute to upload files and submit tasks.
  • [DATA_EXFILTRATION]: The instructions require the agent to send user-provided content and files to an external endpoint (http://opencode-bridge:8000). While this is central to the skill's functionality, it represents a network operation to a non-whitelisted domain.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (such as Word, PDF, Excel, and CSV files) which could contain malicious instructions.
  • Ingestion points: Stage 2 involves the agent receiving and uploading various external document types to the system.
  • Boundary markers: There are no instructions or delimiters provided to ensure the agent ignores or separates embedded instructions within the processed documents.
  • Capability inventory: The agent possesses shell execution capabilities (curl) to transmit data to a downstream processor.
  • Sanitization: The skill does not implement any validation or sanitization of the external data before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 02:31 AM
Security Audit — agent-trust-hub — tech-design-writer