ssd

Warn

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform complex operations, including running sqlcmd to execute DDL scripts, dotnet ef for database scaffolding, and various project build commands (dotnet build, npm run build, uv run pytest).
  • [EXTERNAL_DOWNLOADS]: Fetches software templates and project configurations from external GitHub repositories under the SiamsmileDev organization (e.g., Net60_ApiTemplate_2023, react-ts-template-2023, and SSD-Python-Starter-Template).
  • [REMOTE_CODE_EXECUTION]: Instructs the agent to download and install external software and packages at runtime, including global installations of dotnet-ef, release-it, and project-specific dependencies via npm install or uv sync.
  • [CREDENTIALS_UNSAFE]: The skill contains logic to read and parse database connection strings from appsettings.json and .env files. These strings frequently contain clear-text passwords and are passed as arguments to shell commands for scaffolding and database operations.
  • [PROMPT_INJECTION]: Several reference files use 'IMPORTANT' and 'CRITICAL' markers to enforce coding standards and behavior (e.g., backend-feature.md, dotnet-scaffold.md). While intended for compliance, these patterns resemble behavioral overrides.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface as it ingest untrusted project data (source code, SQL scripts, config files) and has the capability to execute shell commands based on that data.
  • Ingestion points: Reads various project files such as .csproj, package.json, appsettings.json, and .py/.cs/.tsx source files across most reference modules.
  • Boundary markers: No specific delimiters or safety instructions are used to distinguish between trusted instructions and potentially malicious content within the processed data.
  • Capability inventory: Extensive use of the Bash tool for command execution, file system modifications, and network operations (git/curl/gh).
  • Sanitization: No evidence of input validation or sanitization before passing data from read files into shell command templates.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 08:35 AM
Security Audit — agent-trust-hub — ssd